Datasets:

librarian-bots

/

arxiv-metadata-snapshot

like 3

Follow

Librarian Bots 53

2405.14577

Domenic Rosati

Domenic Rosati, Jan Wehner, Kai Williams, {\L}ukasz Bartoszcze, David Atanasov, Robie Gonzales, Subhabrata Majumdar, Carsten Maple, Hassan Sajjad, Frank Rudzicz

Representation Noising: A Defence Mechanism Against Harmful Finetuning

Published in NeurIPs 2024

cs.CL cs.LG

http://creativecommons.org/licenses/by/4.0/

Releasing open-source large language models (LLMs) presents a dual-use risk since bad actors can easily fine-tune these models for harmful purposes. Even without the open release of weights, weight stealing and fine-tuning APIs make closed models vulnerable to harmful fine-tuning attacks (HFAs). While safety measures like preventing jailbreaks and improving safety guardrails are important, such measures can easily be reversed through fine-tuning. In this work, we propose Representation Noising (RepNoise), a defence mechanism that operates even when attackers have access to the weights. RepNoise works by removing information about harmful representations such that it is difficult to recover them during fine-tuning. Importantly, our defence is also able to generalize across different subsets of harm that have not been seen during the defence process as long as they are drawn from the same distribution of the attack set. Our method does not degrade the general capability of LLMs and retains the ability to train the model on harmless tasks. We provide empirical evidence that the efficacy of our defence lies in its ``depth'': the degree to which information about harmful representations is removed across all layers of the LLM. We also find areas where RepNoise still remains ineffective and highlight how those limitations can inform future research.

2405.14664

Oscar Davis

Oscar Davis, Samuel Kessler, Mircea Petrache, \.Ismail \.Ilkan Ceylan, Michael Bronstein, Avishek Joey Bose

Fisher Flow Matching for Generative Modeling over Discrete Data

NeurIPS 2024

cs.LG cs.AI

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Generative modeling over discrete data has recently seen numerous success stories, with applications spanning language modeling, biological sequence design, and graph-structured molecular data. The predominant generative modeling paradigm for discrete data is still autoregressive, with more recent alternatives based on diffusion or flow-matching falling short of their impressive performance in continuous data settings, such as image or video generation. In this work, we introduce Fisher-Flow, a novel flow-matching model for discrete data. Fisher-Flow takes a manifestly geometric perspective by considering categorical distributions over discrete data as points residing on a statistical manifold equipped with its natural Riemannian metric: the $\textit{Fisher-Rao metric}$. As a result, we demonstrate discrete data itself can be continuously reparameterised to points on the positive orthant of the $d$-hypersphere $\mathbb{S}^d_+$, which allows us to define flows that map any source distribution to target in a principled manner by transporting mass along (closed-form) geodesics of $\mathbb{S}^d_+$. Furthermore, the learned flows in Fisher-Flow can be further bootstrapped by leveraging Riemannian optimal transport leading to improved training dynamics. We prove that the gradient flow induced by Fisher-Flow is optimal in reducing the forward KL divergence. We evaluate Fisher-Flow on an array of synthetic and diverse real-world benchmarks, including designing DNA Promoter, and DNA Enhancer sequences. Empirically, we find that Fisher-Flow improves over prior diffusion and flow-matching models on these benchmarks.

2405.14683

Suvo Banik

Suvo Banik, Partha Sarathi Dutta, Sukriti Manna, Subramanian KRS Sankaranarayanan

Development of a Machine Learning Potential to Study Structure and Thermodynamics of Nickel Nanoclusters

cond-mat.mtrl-sci

http://creativecommons.org/licenses/by/4.0/

Machine Learning (ML) potentials such as Gaussian Approximation Potential (GAP) have demonstrated impressive capabilities in mapping structure to properties across diverse systems. Here, we introduce a GAP model for low-dimensional Ni nanoclusters and demonstrate its flexibility and effectiveness in capturing the energetics, structural diversity and thermodynamic properties of Ni nanoclusters across a broad size range. Through a systematic approach encompassing model development, validation, and application, we evaluate the model's efficacy in representing energetics and configurational features in low-dimensional regimes, while also examining its extrapolative nature to vastly different spatiotemporal regimes. Our analysis and discussion shed light on the data quality required to effectively train such models. Trajectories from large scale MD simulations using the GAP model analyzed with data-driven models like Graph Neural Networks (GNN) reveal intriguing insights into the size-dependent phase behavior and thermo-mechanical stability characteristics of porous Ni nanoparticles. Overall, our work underscores the potential of ML models which coupled with data-driven approaches serve as versatile tools for studying low-dimensional systems and complex material dynamics.

2405.14702

Pengyue Jia

Pengyue Jia, Yiding Liu, Xiaopeng Li, Yuhao Wang, Yantong Du, Xiao Han, Xuetao Wei, Shuaiqiang Wang, Dawei Yin, Xiangyu Zhao

G3: An Effective and Adaptive Framework for Worldwide Geolocalization Using Large Multi-Modality Models

Accepted to NeurIPS2024

cs.CV cs.AI

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Worldwide geolocalization aims to locate the precise location at the coordinate level of photos taken anywhere on the Earth. It is very challenging due to 1) the difficulty of capturing subtle location-aware visual semantics, and 2) the heterogeneous geographical distribution of image data. As a result, existing studies have clear limitations when scaled to a worldwide context. They may easily confuse distant images with similar visual contents, or cannot adapt to various locations worldwide with different amounts of relevant data. To resolve these limitations, we propose G3, a novel framework based on Retrieval-Augmented Generation (RAG). In particular, G3 consists of three steps, i.e., Geo-alignment, Geo-diversification, and Geo-verification to optimize both retrieval and generation phases of worldwide geolocalization. During Geo-alignment, our solution jointly learns expressive multi-modal representations for images, GPS and textual descriptions, which allows us to capture location-aware semantics for retrieving nearby images for a given query. During Geo-diversification, we leverage a prompt ensembling method that is robust to inconsistent retrieval performance for different image queries. Finally, we combine both retrieved and generated GPS candidates in Geo-verification for location prediction. Experiments on two well-established datasets IM2GPS3k and YFCC4k verify the superiority of G3 compared to other state-of-the-art methods. Our code and data are available online for reproduction.

2405.14762

Nicholas Gao

Nicholas Gao, Stephan G\"unnemann

Neural Pfaffians: Solving Many Many-Electron Schr\"odinger Equations

cs.LG physics.chem-ph physics.comp-ph quant-ph

http://creativecommons.org/licenses/by/4.0/

Neural wave functions accomplished unprecedented accuracies in approximating the ground state of many-electron systems, though at a high computational cost. Recent works proposed amortizing the cost by learning generalized wave functions across different structures and compounds instead of solving each problem independently. Enforcing the permutation antisymmetry of electrons in such generalized neural wave functions remained challenging as existing methods require discrete orbital selection via non-learnable hand-crafted algorithms. This work tackles the problem by defining overparametrized, fully learnable neural wave functions suitable for generalization across molecules. We achieve this by relying on Pfaffians rather than Slater determinants. The Pfaffian allows us to enforce the antisymmetry on arbitrary electronic systems without any constraint on electronic spin configurations or molecular structure. Our empirical evaluation finds that a single neural Pfaffian calculates the ground state and ionization energies with chemical accuracy across various systems. On the TinyMol dataset, we outperform the `gold-standard' CCSD(T) CBS reference energies by 1.9m$E_h$ and reduce energy errors compared to previous generalized neural wave functions by up to an order of magnitude.

2405.14808

Zhijing Jin

Zhijing Jin, Nils Heil, Jiarui Liu, Shehzaad Dhuliawala, Yahang Qi, Bernhard Sch\"olkopf, Rada Mihalcea, Mrinmaya Sachan

Implicit Personalization in Language Models: A Systematic Study

EMNLP 2024 Findings

cs.CL cs.AI cs.CY cs.HC cs.LG

http://creativecommons.org/licenses/by-sa/4.0/

Implicit Personalization (IP) is a phenomenon of language models inferring a user's background from the implicit cues in the input prompts and tailoring the response based on this inference. While previous work has touched upon various instances of this problem, there lacks a unified framework to study this behavior. This work systematically studies IP through a rigorous mathematical formulation, a multi-perspective moral reasoning framework, and a set of case studies. Our theoretical foundation for IP relies on a structural causal model and introduces a novel method, indirect intervention, to estimate the causal effect of a mediator variable that cannot be directly intervened upon. Beyond the technical approach, we also introduce a set of moral reasoning principles based on three schools of moral philosophy to study when IP may or may not be ethically appropriate. Equipped with both mathematical and ethical insights, we present three diverse case studies illustrating the varied nature of the IP problem and offer recommendations for future research. Our code is at https://github.com/jiarui-liu/IP, and our data is at https://huggingface.co/datasets/Jerry999/ImplicitPersonalizationData.

2405.14992

Corey Zhou

Li Ji-An, Corey Y. Zhou, Marcus K. Benna, Marcelo G. Mattar

Linking In-context Learning in Transformers to Human Episodic Memory

NeurIPS 2024

cs.CL cs.LG

http://creativecommons.org/licenses/by/4.0/

Understanding connections between artificial and biological intelligent systems can reveal fundamental principles of general intelligence. While many artificial intelligence models have a neuroscience counterpart, such connections are largely missing in Transformer models and the self-attention mechanism. Here, we examine the relationship between interacting attention heads and human episodic memory. We focus on induction heads, which contribute to in-context learning in Transformer-based large language models (LLMs). We demonstrate that induction heads are behaviorally, functionally, and mechanistically similar to the contextual maintenance and retrieval (CMR) model of human episodic memory. Our analyses of LLMs pre-trained on extensive text data show that CMR-like heads often emerge in the intermediate and late layers, qualitatively mirroring human memory biases. The ablation of CMR-like heads suggests their causal role in in-context learning. Our findings uncover a parallel between the computational mechanisms of LLMs and human memory, offering valuable insights into both research fields.

2405.15010

Yuki Takezawa

Yuki Takezawa, Han Bao, Ryoma Sato, Kenta Niwa, Makoto Yamada

Parameter-free Clipped Gradient Descent Meets Polyak

NeurIPS 2024

cs.LG math.OC

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Gradient descent and its variants are de facto standard algorithms for training machine learning models. As gradient descent is sensitive to its hyperparameters, we need to tune the hyperparameters carefully using a grid search. However, the method is time-consuming, particularly when multiple hyperparameters exist. Therefore, recent studies have analyzed parameter-free methods that adjust the hyperparameters on the fly. However, the existing work is limited to investigations of parameter-free methods for the stepsize, and parameter-free methods for other hyperparameters have not been explored. For instance, although the gradient clipping threshold is a crucial hyperparameter in addition to the stepsize for preventing gradient explosion issues, none of the existing studies have investigated parameter-free methods for clipped gradient descent. Therefore, in this study, we investigate the parameter-free methods for clipped gradient descent. Specifically, we propose Inexact Polyak Stepsize, which converges to the optimal solution without any hyperparameters tuning, and its convergence rate is asymptotically independent of $L$ under $L$-smooth and $(L_0, L_1)$-smooth assumptions of the loss function, similar to that of clipped gradient descent with well-tuned hyperparameters. We numerically validated our convergence results using a synthetic function and demonstrated the effectiveness of our proposed methods using LSTM, Nano-GPT, and T5.

2405.15035

albert bonnefous

Albert Bonnefous, Yann Brenier, Roya Mohayaee

Monge-Amp\`ere gravity, optimal transport theory and their link to the Galileons

16 pages, 3 figures, Accepted for publication in Physical Review D

gr-qc astro-ph.CO

http://creativecommons.org/licenses/by/4.0/

Mathematicians have been proposing for sometimes that Monge-Amp\`ere equation, a nonlinear generalization of the Poisson equation, where trace of the Hessian is replaced by its determinant, provides an alternative non-relativistic description of gravity. Monge-Amp\`ere equation is affine invariant, has rich geometric properties, connects to optimal transport theory, and remains bounded at short distances. Monge-Amp\`ere gravity, that uses a slightly different form of the Monge-Amp\`ere equation, naturally emerges through the application of large-deviation principle to a Brownian system of indistinguishable and independent particles. In this work we provide a physical formulation of this mathematical model, study its theoretical viability and confront it with observations. We show that Monge-Amp\`ere gravity cannot replace the Newtonian gravity as it does not withstand the solar-system test. We then show that Monge-Amp\`ere gravity can describe a scalar field, often evoked in modified theories of gravity such as Galileons. We show that Monge-Amp\`ere gravity, as a nonlinear model of a new scalar field, is screened at short distances, and behaves differently from Newtonian gravity above galactic scales but approaches it asymptotically. Finally, we write a relativistic Lagrangian for Monge-Amp\`ere gravity in flat space time, which is the field equation of a sum of the Lagrangians of all Galileons. We also show how the Monge-Amp\`ere equation can be obtained from the fully covariant Lagrangian of quartic Galileon in the static limit. The connection between optimal transport theory and modified theories of gravity with second-order field equations, unravelled here, remains a promising domain to further explore.

2405.15071

Boshi Wang

Boshi Wang, Xiang Yue, Yu Su, Huan Sun

Grokked Transformers are Implicit Reasoners: A Mechanistic Journey to the Edge of Generalization

NeurIPS 2024. Code and data: https://github.com/OSU-NLP-Group/GrokkedTransformer

cs.CL

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

We study whether transformers can learn to implicitly reason over parametric knowledge, a skill that even the most capable language models struggle with. Focusing on two representative reasoning types, composition and comparison, we consistently find that transformers can learn implicit reasoning, but only through grokking, i.e., extended training far beyond overfitting. The levels of generalization also vary across reasoning types: when faced with out-of-distribution examples, transformers fail to systematically generalize for composition but succeed for comparison. We delve into the model's internals throughout training, conducting analytical experiments that reveal: 1) the mechanism behind grokking, such as the formation of the generalizing circuit and its relation to the relative efficiency of generalizing and memorizing circuits, and 2) the connection between systematicity and the configuration of the generalizing circuit. Our findings guide data and training setup to better induce implicit reasoning and suggest potential improvements to the transformer architecture, such as encouraging cross-layer knowledge sharing. Furthermore, we demonstrate that for a challenging reasoning task with a large search space, GPT-4-Turbo and Gemini-1.5-Pro based on non-parametric memory fail badly regardless of prompting styles or retrieval augmentation, while a fully grokked transformer can achieve near-perfect accuracy, showcasing the power of parametric memory for complex reasoning.

2405.15223

Jialong Wu

Jialong Wu, Shaofeng Yin, Ningya Feng, Xu He, Dong Li, Jianye Hao, Mingsheng Long

iVideoGPT: Interactive VideoGPTs are Scalable World Models

NeurIPS 2024. Code is available at project website: https://thuml.github.io/iVideoGPT

cs.CV cs.LG cs.RO

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

World models empower model-based agents to interactively explore, reason, and plan within imagined environments for real-world decision-making. However, the high demand for interactivity poses challenges in harnessing recent advancements in video generative models for developing world models at scale. This work introduces Interactive VideoGPT (iVideoGPT), a scalable autoregressive transformer framework that integrates multimodal signals--visual observations, actions, and rewards--into a sequence of tokens, facilitating an interactive experience of agents via next-token prediction. iVideoGPT features a novel compressive tokenization technique that efficiently discretizes high-dimensional visual observations. Leveraging its scalable architecture, we are able to pre-train iVideoGPT on millions of human and robotic manipulation trajectories, establishing a versatile foundation that is adaptable to serve as interactive world models for a wide range of downstream tasks. These include action-conditioned video prediction, visual planning, and model-based reinforcement learning, where iVideoGPT achieves competitive performance compared with state-of-the-art methods. Our work advances the development of interactive general world models, bridging the gap between generative video models and practical model-based reinforcement learning applications. Code and pre-trained models are available at https://thuml.github.io/iVideoGPT.

2405.15688

Ted Lentsch

Ted Lentsch, Holger Caesar, Dariu M. Gavrila

UNION: Unsupervised 3D Object Detection using Object Appearance-based Pseudo-Classes

NeurIPS'24

cs.CV

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Unsupervised 3D object detection methods have emerged to leverage vast amounts of data without requiring manual labels for training. Recent approaches rely on dynamic objects for learning to detect mobile objects but penalize the detections of static instances during training. Multiple rounds of (self) training are used to add detected static instances to the set of training targets; this procedure to improve performance is computationally expensive. To address this, we propose the method UNION. We use spatial clustering and self-supervised scene flow to obtain a set of static and dynamic object proposals from LiDAR. Subsequently, object proposals' visual appearances are encoded to distinguish static objects in the foreground and background by selecting static instances that are visually similar to dynamic objects. As a result, static and dynamic mobile objects are obtained together, and existing detectors can be trained with a single training. In addition, we extend 3D object discovery to detection by using object appearance-based cluster labels as pseudo-class labels for training object classification. We conduct extensive experiments on the nuScenes dataset and increase the state-of-the-art performance for unsupervised 3D object discovery, i.e. UNION more than doubles the average precision to 38.4. The code is available at github.com/TedLentsch/UNION.

2405.15743

Nolan Dey

Nolan Dey and Shane Bergsma and Joel Hestness

Sparse maximal update parameterization: A holistic approach to sparse training dynamics

10 pages main text, 10 pages reference and appendix, 14 figures, NeurIPS Camera-Ready

cs.LG

http://creativecommons.org/licenses/by/4.0/

Several challenges make it difficult for sparse neural networks to compete with dense models. First, setting a large fraction of weights to zero impairs forward and gradient signal propagation. Second, sparse studies often need to test multiple sparsity levels, while also introducing new hyperparameters (HPs), leading to prohibitive tuning costs. Indeed, the standard practice is to re-use the learning HPs originally crafted for dense models. Unfortunately, we show sparse and dense networks do not share the same optimal HPs. Without stable dynamics and effective training recipes, it is costly to test sparsity at scale, which is key to surpassing dense networks and making the business case for sparsity acceleration in hardware. A holistic approach is needed to tackle these challenges and we propose S$\mu$Par as one such approach. For random unstructured static sparsity, S$\mu$Par ensures activations, gradients, and weight updates all scale independently of sparsity level. Further, by reparameterizing the HPs, S$\mu$Par enables the same HP values to be optimal as we vary both sparsity level and model width. HPs can be tuned on small dense networks and transferred to large sparse models, greatly reducing tuning costs. On large-scale language modeling, S$\mu$Par shows increasing improvements over standard parameterization as sparsity increases, leading up to 11.9% relative loss improvement at 99.2% sparsity. A minimal implementation of S$\mu$Par is available at https://github.com/EleutherAI/nanoGPT-mup/tree/supar.

2405.16173

Shutong Ding

Shutong Ding, Ke Hu, Zhenhao Zhang, Kan Ren, Weinan Zhang, Jingyi Yu, Jingya Wang, Ye Shi

Diffusion-based Reinforcement Learning via Q-weighted Variational Policy Optimization

Accepted by NeurIPS2024

cs.LG

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Diffusion models have garnered widespread attention in Reinforcement Learning (RL) for their powerful expressiveness and multimodality. It has been verified that utilizing diffusion policies can significantly improve the performance of RL algorithms in continuous control tasks by overcoming the limitations of unimodal policies, such as Gaussian policies, and providing the agent with enhanced exploration capabilities. However, existing works mainly focus on the application of diffusion policies in offline RL, while their incorporation into online RL is less investigated. The training objective of the diffusion model, known as the variational lower bound, cannot be optimized directly in online RL due to the unavailability of 'good' actions. This leads to difficulties in conducting diffusion policy improvement. To overcome this, we propose a novel model-free diffusion-based online RL algorithm, Q-weighted Variational Policy Optimization (QVPO). Specifically, we introduce the Q-weighted variational loss, which can be proved to be a tight lower bound of the policy objective in online RL under certain conditions. To fulfill these conditions, the Q-weight transformation functions are introduced for general scenarios. Additionally, to further enhance the exploration capability of the diffusion policy, we design a special entropy regularization term. We also develop an efficient behavior policy to enhance sample efficiency by reducing the variance of the diffusion policy during online interactions. Consequently, the QVPO algorithm leverages the exploration capabilities and multimodality of diffusion policies, preventing the RL agent from converging to a sub-optimal policy. To verify the effectiveness of QVPO, we conduct comprehensive experiments on MuJoCo benchmarks. The final results demonstrate that QVPO achieves state-of-the-art performance on both cumulative reward and sample efficiency.

2405.16876

Yidong Ouyang

Yidong Ouyang, Liyan Xie, Hongyuan Zha, Guang Cheng

Transfer Learning for Diffusion Models

NeurIPS 2024

cs.LG cs.AI

http://creativecommons.org/licenses/by/4.0/

Diffusion models, a specific type of generative model, have achieved unprecedented performance in recent years and consistently produce high-quality synthetic samples. A critical prerequisite for their notable success lies in the presence of a substantial number of training samples, which can be impractical in real-world applications due to high collection costs or associated risks. Consequently, various finetuning and regularization approaches have been proposed to transfer knowledge from existing pre-trained models to specific target domains with limited data. This paper introduces the Transfer Guided Diffusion Process (TGDP), a novel approach distinct from conventional finetuning and regularization methods. We prove that the optimal diffusion model for the target domain integrates pre-trained diffusion models on the source domain with additional guidance from a domain classifier. We further extend TGDP to a conditional version for modeling the joint distribution of data and its corresponding labels, together with two additional regularization terms to enhance the model performance. We validate the effectiveness of TGDP on both simulated and real-world datasets.

2405.16978

Yuefeng Peng

Yuefeng Peng, Jaechul Roh, Subhransu Maji and Amir Houmansadr

OSLO: One-Shot Label-Only Membership Inference Attacks

To appear at NeurIPS 2024

cs.LG cs.CR

http://creativecommons.org/licenses/by/4.0/

We introduce One-Shot Label-Only (OSLO) membership inference attacks (MIAs), which accurately infer a given sample's membership in a target model's training set with high precision using just \emph{a single query}, where the target model only returns the predicted hard label. This is in contrast to state-of-the-art label-only attacks which require $\sim6000$ queries, yet get attack precisions lower than OSLO's. OSLO leverages transfer-based black-box adversarial attacks. The core idea is that a member sample exhibits more resistance to adversarial perturbations than a non-member. We compare OSLO against state-of-the-art label-only attacks and demonstrate that, despite requiring only one query, our method significantly outperforms previous attacks in terms of precision and true positive rate (TPR) under the same false positive rates (FPR). For example, compared to previous label-only MIAs, OSLO achieves a TPR that is at least 7$\times$ higher under a 1\% FPR and at least 22$\times$ higher under a 0.1\% FPR on CIFAR100 for a ResNet18 model. We evaluated multiple defense mechanisms against OSLO.

2405.17075

Jia-Jie Zhu

Egor Gladin, Pavel Dvurechensky, Alexander Mielke, Jia-Jie Zhu

Interaction-Force Transport Gradient Flows

cs.LG math.AP stat.ML

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

This paper presents a new gradient flow dissipation geometry over non-negative and probability measures. This is motivated by a principled construction that combines the unbalanced optimal transport and interaction forces modeled by reproducing kernels. Using a precise connection between the Hellinger geometry and the maximum mean discrepancy (MMD), we propose the interaction-force transport (IFT) gradient flows and its spherical variant via an infimal convolution of the Wasserstein and spherical MMD tensors. We then develop a particle-based optimization algorithm based on the JKO-splitting scheme of the mass-preserving spherical IFT gradient flows. Finally, we provide both theoretical global exponential convergence guarantees and improved empirical simulation results for applying the IFT gradient flows to the sampling task of MMD-minimization. Furthermore, we prove that the spherical IFT gradient flow enjoys the best of both worlds by providing the global exponential convergence guarantee for both the MMD and KL energy.

2405.17151

Riccardo Cadei

Riccardo Cadei, Lukas Lindorfer, Sylvia Cremer, Cordelia Schmid, and Francesco Locatello

Smoke and Mirrors in Causal Downstream Tasks

cs.LG

http://creativecommons.org/licenses/by/4.0/

Machine Learning and AI have the potential to transform data-driven scientific discovery, enabling accurate predictions for several scientific phenomena. As many scientific questions are inherently causal, this paper looks at the causal inference task of treatment effect estimation, where the outcome of interest is recorded in high-dimensional observations in a Randomized Controlled Trial (RCT). Despite being the simplest possible causal setting and a perfect fit for deep learning, we theoretically find that many common choices in the literature may lead to biased estimates. To test the practical impact of these considerations, we recorded ISTAnt, the first real-world benchmark for causal inference downstream tasks on high-dimensional observations as an RCT studying how garden ants (Lasius neglectus) respond to microparticles applied onto their colony members by hygienic grooming. Comparing 6 480 models fine-tuned from state-of-the-art visual backbones, we find that the sampling and modeling choices significantly affect the accuracy of the causal estimate, and that classification accuracy is not a proxy thereof. We further validated the analysis, repeating it on a synthetically generated visual data set controlling the causal model. Our results suggest that future benchmarks should carefully consider real downstream scientific questions, especially causal ones. Further, we highlight guidelines for representation learning methods to help answer causal questions in the sciences.

2405.17202

Felipe Maia Polo

Felipe Maia Polo, Ronald Xu, Lucas Weber, M\'irian Silva, Onkar Bhardwaj, Leshem Choshen, Allysson Flavio Melo de Oliveira, Yuekai Sun, Mikhail Yurochkin

Efficient multi-prompt evaluation of LLMs

NeurIPS 2024

cs.CL cs.AI cs.LG stat.ML

http://creativecommons.org/licenses/by/4.0/

Most popular benchmarks for comparing LLMs rely on a limited set of prompt templates, which may not fully capture the LLMs' abilities and can affect the reproducibility of results on leaderboards. Many recent works empirically verify prompt sensitivity and advocate for changes in LLM evaluation. In this paper, we consider the problem of estimating the performance distribution across many prompt variants instead of finding a single prompt to evaluate with. We introduce PromptEval, a method for estimating performance across a large set of prompts borrowing strength across prompts and examples to produce accurate estimates under practical evaluation budgets. The resulting distribution can be used to obtain performance quantiles to construct various robust performance metrics (e.g., top 95% quantile or median). We prove that PromptEval consistently estimates the performance distribution and demonstrate its efficacy empirically on three prominent LLM benchmarks: MMLU, BIG-bench Hard, and LMentry; for example, PromptEval can accurately estimate performance quantiles across 100 prompt templates on MMLU with a budget equivalent to two single-prompt evaluations. Moreover, we show how PromptEval can be useful in LLM-as-a-judge and best prompt identification applications.

2405.17374

ShengYun Peng

ShengYun Peng, Pin-Yu Chen, Matthew Hull, Duen Horng Chau

Navigating the Safety Landscape: Measuring Risks in Finetuning Large Language Models

NeurIPS'24

cs.LG

http://creativecommons.org/licenses/by/4.0/

Safety alignment is crucial to ensure that large language models (LLMs) behave in ways that align with human preferences and prevent harmful actions during inference. However, recent studies show that the alignment can be easily compromised through finetuning with only a few adversarially designed training examples. We aim to measure the risks in finetuning LLMs through navigating the LLM safety landscape. We discover a new phenomenon observed universally in the model parameter space of popular open-source LLMs, termed as "safety basin": random perturbations to model weights maintain the safety level of the original aligned model within its local neighborhood. However, outside this local region, safety is fully compromised, exhibiting a sharp, step-like drop. This safety basin contrasts sharply with the LLM capability landscape, where model performance peaks at the origin and gradually declines as random perturbation increases. Our discovery inspires us to propose the new VISAGE safety metric that measures the safety in LLM finetuning by probing its safety landscape. Visualizing the safety landscape of the aligned model enables us to understand how finetuning compromises safety by dragging the model away from the safety basin. The LLM safety landscape also highlights the system prompt's critical role in protecting a model, and that such protection transfers to its perturbed variants within the safety basin. These observations from our safety landscape research provide new insights for future work on LLM safety community. Our code is publicly available at https://github.com/ShengYun-Peng/llm-landscape.

2405.17700

Itai Shapira

Kanad Shrikar Pardeshi, Itai Shapira, Ariel D. Procaccia, Aarti Singh

Learning Social Welfare Functions

cs.GT cs.LG

http://creativecommons.org/licenses/by/4.0/

Is it possible to understand or imitate a policy maker's rationale by looking at past decisions they made? We formalize this question as the problem of learning social welfare functions belonging to the well-studied family of power mean functions. We focus on two learning tasks; in the first, the input is vectors of utilities of an action (decision or policy) for individuals in a group and their associated social welfare as judged by a policy maker, whereas in the second, the input is pairwise comparisons between the welfares associated with a given pair of utility vectors. We show that power mean functions are learnable with polynomial sample complexity in both cases, even if the comparisons are social welfare information is noisy. Finally, we design practical algorithms for these tasks and evaluate their performance.

2405.17809

Chenyang Le

Chenyang Le, Yao Qian, Dongmei Wang, Long Zhou, Shujie Liu, Xiaofei Wang, Midia Yousefi, Yanmin Qian, Jinyu Li, Sheng Zhao, Michael Zeng

TransVIP: Speech to Speech Translation System with Voice and Isochrony Preservation

Neural Information Processing Systems, poster

cs.CL cs.AI cs.SD eess.AS

http://creativecommons.org/licenses/by-nc-nd/4.0/

There is a rising interest and trend in research towards directly translating speech from one language to another, known as end-to-end speech-to-speech translation. However, most end-to-end models struggle to outperform cascade models, i.e., a pipeline framework by concatenating speech recognition, machine translation and text-to-speech models. The primary challenges stem from the inherent complexities involved in direct translation tasks and the scarcity of data. In this study, we introduce a novel model framework TransVIP that leverages diverse datasets in a cascade fashion yet facilitates end-to-end inference through joint probability. Furthermore, we propose two separated encoders to preserve the speaker's voice characteristics and isochrony from the source speech during the translation process, making it highly suitable for scenarios such as video dubbing. Our experiments on the French-English language pair demonstrate that our model outperforms the current state-of-the-art speech-to-speech translation model.

2405.17983

Dean Brandner

Dean Brandner and Sergio Lucia

Reinforced Model Predictive Control via Trust-Region Quasi-Newton Policy Optimization

European Control Conference (ECC) 2024, Code: https://github.com/DeanBrandner/ECC24_TR_improved_QN_PO_for_MPC_in_RL

10.23919/ECC64448.2024.10590970

cs.LG cs.SY eess.SY

http://creativecommons.org/licenses/by/4.0/

Model predictive control can optimally deal with nonlinear systems under consideration of constraints. The control performance depends on the model accuracy and the prediction horizon. Recent advances propose to use reinforcement learning applied to a parameterized model predictive controller to recover the optimal control performance even if an imperfect model or short prediction horizons are used. However, common reinforcement learning algorithms rely on first order updates, which only have a linear convergence rate and hence need an excessive amount of dynamic data. Higher order updates are typically intractable if the policy is approximated with neural networks due to the large number of parameters. In this work, we use a parameterized model predictive controller as policy, and leverage the small amount of necessary parameters to propose a trust-region constrained Quasi-Newton training algorithm for policy optimization with a superlinear convergence rate. We show that the required second order derivative information can be calculated by the solution of a linear system of equations. A simulation study illustrates that the proposed training algorithm outperforms other algorithms in terms of data efficiency and accuracy.

2405.18400

Ethan Shen

Ethan Shen, Alan Fan, Sarah M. Pratt, Jae Sung Park, Matthew Wallingford, Sham M. Kakade, Ari Holtzman, Ranjay Krishna, Ali Farhadi, Aditya Kusupati

Superposed Decoding: Multiple Generations from a Single Autoregressive Inference Pass

23 pages, 16 figures, accepted at NeurIPS 2024

cs.CL cs.LG

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Many applications today provide users with multiple auto-complete drafts as they type, including GitHub's code completion, Gmail's smart compose, and Apple's messaging auto-suggestions. Under the hood, language models support this by running an autoregressive inference pass to provide a draft. Consequently, providing $k$ drafts to the user requires running an expensive language model $k$ times. To alleviate the computation cost of running $k$ inference passes, we propose Superposed Decoding, a new decoding algorithm that generates $k$ drafts at the computation cost of one autoregressive inference pass. We achieve this by feeding a superposition of the most recent token embeddings from the $k$ drafts as input to the next decoding step of the language model. At every inference step we combine the $k$ drafts with the top-$k$ tokens to get $k^2$ new drafts and cache the $k$ most likely options, using an n-gram interpolation with minimal compute overhead to filter out incoherent generations. Our experiments show that $k$ drafts from Superposed Decoding are at least as coherent and factual as Nucleus Sampling and Greedy Decoding respectively, while being at least $2.44\times$ faster for $k\ge3$. In a compute-normalized setting, user evaluations demonstrably favor text generated by Superposed Decoding over Nucleus Sampling. Superposed Decoding can also be combined with other decoding strategies, resulting in universal coverage gains when scaling inference time compute. Code and more examples open-sourced at https://github.com/RAIVNLab/SuperposedDecoding.

2405.18457

Jihao Andreas Lin

Jihao Andreas Lin and Shreyas Padhy and Bruno Mlodozeniec and Javier Antor\'an and Jos\'e Miguel Hern\'andez-Lobato

Improving Linear System Solvers for Hyperparameter Optimisation in Iterative Gaussian Processes

Advances in Neural Information Processing Systems 2024

cs.LG stat.ML

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Scaling hyperparameter optimisation to very large datasets remains an open problem in the Gaussian process community. This paper focuses on iterative methods, which use linear system solvers, like conjugate gradients, alternating projections or stochastic gradient descent, to construct an estimate of the marginal likelihood gradient. We discuss three key improvements which are applicable across solvers: (i) a pathwise gradient estimator, which reduces the required number of solver iterations and amortises the computational cost of making predictions, (ii) warm starting linear system solvers with the solution from the previous step, which leads to faster solver convergence at the cost of negligible bias, (iii) early stopping linear system solvers after a limited computational budget, which synergises with warm starting, allowing solver progress to accumulate over multiple marginal likelihood steps. These techniques provide speed-ups of up to $72\times$ when solving to tolerance, and decrease the average residual norm by up to $7\times$ when stopping early.

2405.18958

L\'eonard Brice

L\'eonard Brice, Jean-Fran\c{c}ois Raskin, Mathieu Sassolas, Guillaume Scerri, Marie van den Bogaard

Pessimism of the Will, Optimism of the Intellect: Fair Protocols with Malicious but Rational Agents

53 pages, 14 figures

cs.GT cs.CC cs.CR

http://creativecommons.org/licenses/by/4.0/

Fairness is a desirable and crucial property of many protocols that handle, for instance, exchanges of message. It states that if at least one agent engaging in the protocol is honest, then either the protocol will unfold correctly and fulfill its intended goal for all participants, or it will fail for everyone. In this work, we present a game-based framework for the study of fairness protocols, that does not define a priori an attacker model. It is based on the notion of strong secure equilibria, and leverages the conceptual and algorithmic toolbox of game theory. In the case of finite games, we provide decision procedures with tight complexity bounds for determining whether a protocol is immune to nefarious attacks from a coalition of participants, and whether such a protocol could exist based on the underlying graph structure and objectives.

2405.19073

Celestine Mendler-D\"unner

Celestine Mendler-D\"unner, Gabriele Carovano, Moritz Hardt

An engine not a camera: Measuring performative power of online search

to appear at NeurIPS 2024

cs.CY cs.IR

http://creativecommons.org/licenses/by-nc-nd/4.0/

The power of digital platforms is at the center of major ongoing policy and regulatory efforts. To advance existing debates, we designed and executed an experiment to measure the performative power of online search providers. Instantiated in our setting, performative power quantifies the ability of a search engine to steer web traffic by rearranging results. To operationalize this definition we developed a browser extension that performs unassuming randomized experiments in the background. These randomized experiments emulate updates to the search algorithm and identify the causal effect of different content arrangements on clicks. Analyzing tens of thousands of clicks, we discuss what our robust quantitative findings say about the power of online search engines, using the Google Shopping antitrust investigation as a case study. More broadly, we envision our work to serve as a blueprint for how the recent definition of performative power can help integrate quantitative insights from online experiments with future investigations into the economic power of digital platforms.

2405.19291

Yi-Lin Wei

Yi-Lin Wei, Jian-Jian Jiang, Chengyi Xing, Xian-Tuo Tan, Xiao-Ming Wu, Hao Li, Mark Cutkosky, Wei-Shi Zheng

Grasp as You Say: Language-guided Dexterous Grasp Generation

Accepted by NeurIPS2024

cs.RO

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

This paper explores a novel task "Dexterous Grasp as You Say" (DexGYS), enabling robots to perform dexterous grasping based on human commands expressed in natural language. However, the development of this field is hindered by the lack of datasets with natural human guidance; thus, we propose a language-guided dexterous grasp dataset, named DexGYSNet, offering high-quality dexterous grasp annotations along with flexible and fine-grained human language guidance. Our dataset construction is cost-efficient, with the carefully-design hand-object interaction retargeting strategy, and the LLM-assisted language guidance annotation system. Equipped with this dataset, we introduce the DexGYSGrasp framework for generating dexterous grasps based on human language instructions, with the capability of producing grasps that are intent-aligned, high quality and diversity. To achieve this capability, our framework decomposes the complex learning process into two manageable progressive objectives and introduce two components to realize them. The first component learns the grasp distribution focusing on intention alignment and generation diversity. And the second component refines the grasp quality while maintaining intention consistency. Extensive experiments are conducted on DexGYSNet and real world environments for validation.

2405.19446

Erica Bertolini

Erica Bertolini, Alberto Blasi, Nicola Maggiore, Daniel Sacco Shaikh

Hall-like behaviour of higher rank Chern-Simons theory of fractons

31 pages, appendix and references added, version to appear on JHEP

JHEP 10 (2024), 232

10.1007/JHEP10(2024)232

DIAS-STP-24-12

hep-th cond-mat.mes-hall cond-mat.str-el gr-qc

http://creativecommons.org/licenses/by/4.0/

Fracton phases of matter constitute an interesting point of contact between condensed matter and high-energy physics. The limited mobility property of fracton quasiparticles finds applications in many different contexts, including quantum information, spin liquids, elasticity, hydrodynamics, gravity and holography. In this paper we adopt a field theoretical approach to investigate the three dimensional action of a rank-2 symmetric tensor field invariant under the covariant fracton symmetry. The theory appears as a non-topological higher rank generalization of the ordinary Chern-Simons model, depending only on the traceless part of the tensor gauge field. After defining a field strength, a rank-2 traceless ``electric'' field and a ``magnetic'' vector field are identified, in analogy with the standard Chern-Simons ones. Once matter is introduced, a Hall-like behaviour with fractonic features emerges. In particular, our model shows a Hall-like dipole current, together with a vectorial ``flux-attachment'' relation for dipoles. This gives a possible starting point for a fracton - vortex duality. A gauge-fixing term is then introduced, from which propagators are computed and the counting of the degrees of freedom is performed. Finally, the energy-momentum tensor is shown to be conserved and the integrated energy density is proved to be zero, which reminds the topological nature of the standard Chern-Simons model.

2405.19510

Qier An

Qier An, Tiger Lu, G.Mirek Brandt, Timothy D Brandt, Gongjie Li

Significant mutual inclinations between the stellar spin and the orbits of both planets in the HAT-P-11 system

14 pages, 10 figures, submitted to AJ

astro-ph.EP

http://creativecommons.org/licenses/by/4.0/

Planet-star obliquity and planet-planet ]mutual inclination encode a planetary system's dynamical history, but both of their values are hard to measure for misaligned systems with close-in companions. HAT-P-11 is a K4 star with two known planets: a close-in, misaligned super-Neptune with a $\approx$5-day orbit, and an outer super-Jupiter with a $\approx$10-year orbit. In this work we present a joint orbit fit of the HAT-P-11 system with astrometry and S-index corrected RV data. By combining our results with previous constraints on the orientation of the star and the inner planet, we find that all three angular momenta -- those of the star, planet b, and planet c -- are significantly misaligned. We confirm the status of planet c as a super-Jupiter, with $2.68\pm0.41\, \mathrm{M_{\rm Jup}}$ at a semimajor axis of $4.10\pm0.06\,$AU, and planet b's mass of $\mathrm{M_b\sin{i_b}}=0.074\pm0.004\, \mathrm{M_{\rm Jup}}$. We present the posterior probability distribution of obliquity between star A and planet c, and mutual inclination between planet b and planet c.

2405.19534

Angelica Chen

Angelica Chen, Sadhika Malladi, Lily H. Zhang, Xinyi Chen, Qiuyi Zhang, Rajesh Ranganath, Kyunghyun Cho

Preference Learning Algorithms Do Not Learn Preference Rankings

NeurIPS 2024 camera-ready

cs.LG cs.AI cs.CL

http://creativecommons.org/licenses/by/4.0/

Preference learning algorithms (e.g., RLHF and DPO) are frequently used to steer LLMs to produce generations that are more preferred by humans, but our understanding of their inner workings is still limited. In this work, we study the conventional wisdom that preference learning trains models to assign higher likelihoods to more preferred outputs than less preferred outputs, measured via ranking accuracy. Surprisingly, we find that most state-of-the-art preference-tuned models achieve a ranking accuracy of less than 60% on common preference datasets. We furthermore derive the idealized ranking accuracy that a preference-tuned LLM would achieve if it optimized the DPO or RLHF objective perfectly. We demonstrate that existing models exhibit a significant alignment gap -- i.e., a gap between the observed and idealized ranking accuracies. We attribute this discrepancy to the DPO objective, which is empirically and theoretically ill-suited to fix even mild ranking errors in the reference model, and derive a simple and efficient formula for quantifying the difficulty of learning a given preference datapoint. Finally, we demonstrate that ranking accuracy strongly correlates with the empirically popular win rate metric when the model is close to the reference model used in the objective, shedding further light on the differences between on-policy (e.g., RLHF) and off-policy (e.g., DPO) preference learning algorithms.

2405.19681

Matt Jones

Matt Jones, Peter Chang, Kevin Murphy

Bayesian Online Natural Gradient (BONG)

Final NeurIPS version, updated in response to reviews. 43 pages, 13 figures

NeurIPS 2024

stat.ML cs.LG stat.CO

http://creativecommons.org/licenses/by/4.0/

We propose a novel approach to sequential Bayesian inference based on variational Bayes (VB). The key insight is that, in the online setting, we do not need to add the KL term to regularize to the prior (which comes from the posterior at the previous timestep); instead we can optimize just the expected log-likelihood, performing a single step of natural gradient descent starting at the prior predictive. We prove this method recovers exact Bayesian inference if the model is conjugate. We also show how to compute an efficient deterministic approximation to the VB objective, as well as our simplified objective, when the variational distribution is Gaussian or a sub-family, including the case of a diagonal plus low-rank precision matrix. We show empirically that our method outperforms other online VB methods in the non-conjugate setting, such as online learning for neural networks, especially when controlling for computational costs.

2405.20039

Jiacheng Miao

Jiacheng Miao and Qiongshi Lu

Task-Agnostic Machine-Learning-Assisted Inference

stat.ML cs.LG stat.ME

http://creativecommons.org/licenses/by/4.0/

Machine learning (ML) is playing an increasingly important role in scientific research. In conjunction with classical statistical approaches, ML-assisted analytical strategies have shown great promise in accelerating research findings. This has also opened a whole field of methodological research focusing on integrative approaches that leverage both ML and statistics to tackle data science challenges. One type of study that has quickly gained popularity employs ML to predict unobserved outcomes in massive samples, and then uses predicted outcomes in downstream statistical inference. However, existing methods designed to ensure the validity of this type of post-prediction inference are limited to very basic tasks such as linear regression analysis. This is because any extension of these approaches to new, more sophisticated statistical tasks requires task-specific algebraic derivations and software implementations, which ignores the massive library of existing software tools already developed for the same scientific problem given observed data. This severely constrains the scope of application for post-prediction inference. To address this challenge, we introduce a novel statistical framework named PSPS for task-agnostic ML-assisted inference. It provides a post-prediction inference solution that can be easily plugged into almost any established data analysis routines. It delivers valid and efficient inference that is robust to arbitrary choice of ML model, allowing nearly all existing statistical frameworks to be incorporated into the analysis of ML-predicted data. Through extensive experiments, we showcase our method's validity, versatility, and superiority compared to existing approaches. Our software is available at https://github.com/qlu-lab/psps.

2405.20165

Taehyun Hwang

Wooseong Cho, Taehyun Hwang, Joongkyu Lee and Min-hwan Oh

Randomized Exploration for Reinforcement Learning with Multinomial Logistic Function Approximation

Accepted in NeurIPS 2024

stat.ML cs.LG

http://creativecommons.org/licenses/by-nc-nd/4.0/

We study reinforcement learning with multinomial logistic (MNL) function approximation where the underlying transition probability kernel of the Markov decision processes (MDPs) is parametrized by an unknown transition core with features of state and action. For the finite horizon episodic setting with inhomogeneous state transitions, we propose provably efficient algorithms with randomized exploration having frequentist regret guarantees. For our first algorithm, $\texttt{RRL-MNL}$, we adapt optimistic sampling to ensure the optimism of the estimated value function with sufficient frequency. We establish that $\texttt{RRL-MNL}$ achieves a $\tilde{O}(\kappa^{-1} d^{\frac{3}{2}} H^{\frac{3}{2}} \sqrt{T})$ frequentist regret bound with constant-time computational cost per episode. Here, $d$ is the dimension of the transition core, $H$ is the horizon length, $T$ is the total number of steps, and $\kappa$ is a problem-dependent constant. Despite the simplicity and practicality of $\texttt{RRL-MNL}$, its regret bound scales with $\kappa^{-1}$, which is potentially large in the worst case. To improve the dependence on $\kappa^{-1}$, we propose $\texttt{ORRL-MNL}$, which estimates the value function using the local gradient information of the MNL transition model. We show that its frequentist regret bound is $\tilde{O}(d^{\frac{3}{2}} H^{\frac{3}{2}} \sqrt{T} + \kappa^{-1} d^2 H^2)$. To the best of our knowledge, these are the first randomized RL algorithms for the MNL transition model that achieve statistical guarantees with constant-time computational cost per episode. Numerical experiments demonstrate the superior performance of the proposed algorithms.

2405.20280

Xiaojun Yao

Di-Lun Yang and Xiaojun Yao

Quarkonium Polarization in Medium from Open Quantum Systems and Chromomagnetic Correlators

38 pages; v2: published version

Phys.Rev.D 110, 074037 (2024)

10.1103/PhysRevD.110.074037

IQuS@UW-21-079

hep-ph nucl-th quant-ph

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

We study the spin-dependent in-medium dynamics of quarkonia by using the potential nonrelativistic QCD (pNRQCD) and the open quantum system framework. We consider the pNRQCD Lagrangian valid up to the order $\frac{r}{M^0}=r$ and $\frac{r^0}{M}=\frac{1}{M}$ in the double power counting. By considering the Markovian condition and applying the Wigner transformation upon the diagonal spin components of the quarkonium density matrix with the semiclassical expansion, we systematically derive the Boltzmann transport equation for quarkonia with polarization dependence in the quantum optical limit. Unlike the spin-independent collision terms governed by certain chromoelectric field correlators, new gauge invariant correlators of chromomagnetic fields determine the recombination and dissociation terms with polarization dependence at the order we are working. We also derive a Lindblad equation describing the in-medium transitions between spin-singlet and spin-triplet heavy quark-antiquark pairs in the quantum Brownian motion limit. The Lindblad equation is governed by new transport coefficients defined in terms of the chromomagnetic field correlators. Our formalism is generic and valid for both weakly coupled and strongly coupled quark gluon plasmas. It can be further applied to study spin alignment of vector quarkonia in heavy ion collisions.

2405.20630

Byoungwoo Park

Byoungwoo Park, Jungwon Choi, Sungbin Lim, Juho Lee

Stochastic Optimal Control for Diffusion Bridges in Function Spaces

cs.LG

http://creativecommons.org/licenses/by/4.0/

Recent advancements in diffusion models and diffusion bridges primarily focus on finite-dimensional spaces, yet many real-world problems necessitate operations in infinite-dimensional function spaces for more natural and interpretable formulations. In this paper, we present a theory of stochastic optimal control (SOC) tailored to infinite-dimensional spaces, aiming to extend diffusion-based algorithms to function spaces. Specifically, we demonstrate how Doob's $h$-transform, the fundamental tool for constructing diffusion bridges, can be derived from the SOC perspective and expanded to infinite dimensions. This expansion presents a challenge, as infinite-dimensional spaces typically lack closed-form densities. Leveraging our theory, we establish that solving the optimal control problem with a specific objective function choice is equivalent to learning diffusion-based generative models. We propose two applications: (1) learning bridges between two infinite-dimensional distributions and (2) generative models for sampling from an infinite-dimensional distribution. Our approach proves effective for diverse problems involving continuous function space representations, such as resolution-free images, time-series data, and probability density functions.

2406.00482

In\'es Pastor-Marazuela Dr.

In\'es Pastor-Marazuela, Joeri van Leeuwen, Anna Bilous, Liam Connor, Yogesh Maan, Leon Oostrum, Emily Petroff, Dany Vohl, Kelley M. Hess, Emanuela Orr\`u, Alessio Sclocco, and Yuyang Wang

Comprehensive analysis of the Apertif Fast Radio Burst sample: similarities with young, energetic neutron stars

40 pages (including 11 of appendix), 37 figures, 5 tables. Accepted for publication in A&A

astro-ph.HE

http://creativecommons.org/licenses/by/4.0/

Understanding the origin of fast radio bursts (FRBs) has become the main science driver of recent dedicated FRB surveys. Between July 2019 and February 2022, we carried out ALERT, an FRB survey at 1370 MHz using the Apertif instrument installed at the Westerbork Synthesis Radio Telescope (WSRT). Here we report the detection of 18 new FRBs, and we study the properties of the entire 24 burst sample detected during the survey. For five bursts, we identify host galaxy candidates with >50% probability association. We observe an average linear polarisation fraction of $\sim$43% and an average circular polarisation fraction consistent with 0%. A third of the FRBs display multiple components. The sample next reveals a population of highly scattered bursts, which is most likely to have been produced in the immediate circumburst environment. Furthermore, two FRBs show evidence for high rotation measures, reaching |RM|>$10^3$ rad m$^{-2}$ in the source reference frames. Together, the scattering and rotation measures ALERT finds prove that a large fraction of FRBs are embedded in complex media such as star forming regions or supernova remnants. Through the discovery of the third most dispersed FRB so far, we show that one-off FRBs can emit at frequencies in excess of 6 GHz. Finally, we determine an FRB all-sky rate of $459^{+208}_{-155}$ sky$^{-1}$ day$^{-1}$ above a fluence limit of 4.1 Jy ms, and a fluence cumulative distribution with a power law index $\gamma=-1.23\pm0.06\pm0.2$, which is roughly consistent with the Euclidean Universe predictions. Through the high resolution in time, frequency, polarisation and localisation that ALERT featured, we were able to determine the morphological complexity, polarisation, local scattering and magnetic environment, and high-frequency luminosity of FRBs. We find all these strongly resemble those seen in young, energetic, highly magnetised neutron stars.

2406.00885

Ivan Moskalenko

Ivan Moskalenko, Anastasiia Kornilova, Gonzalo Ferrer

Visual place recognition for aerial imagery: A survey

Robotics and Autonomous Systems 183 (2025) 104837

10.1016/j.robot.2024.104837

cs.CV cs.RO

http://creativecommons.org/licenses/by-nc-sa/4.0/

Aerial imagery and its direct application to visual localization is an essential problem for many Robotics and Computer Vision tasks. While Global Navigation Satellite Systems (GNSS) are the standard default solution for solving the aerial localization problem, it is subject to a number of limitations, such as, signal instability or solution unreliability that make this option not so desirable. Consequently, visual geolocalization is emerging as a viable alternative. However, adapting Visual Place Recognition (VPR) task to aerial imagery presents significant challenges, including weather variations and repetitive patterns. Current VPR reviews largely neglect the specific context of aerial data. This paper introduces a methodology tailored for evaluating VPR techniques specifically in the domain of aerial imagery, providing a comprehensive assessment of various methods and their performance. However, we not only compare various VPR methods, but also demonstrate the importance of selecting appropriate zoom and overlap levels when constructing map tiles to achieve maximum efficiency of VPR algorithms in the case of aerial imagery. The code is available on our GitHub repository -- https://github.com/prime-slam/aero-vloc.

2406.00985

Jialing Cai

Mingzhen Huang, Jialing Cai, Shan Jia, Vishnu Suresh Lokhande, Siwei Lyu

ParallelEdits: Efficient Multi-Aspect Text-Driven Image Editing with Attention Grouping

cs.CV

http://creativecommons.org/licenses/by/4.0/

Text-driven image synthesis has made significant advancements with the development of diffusion models, transforming how visual content is generated from text prompts. Despite these advances, text-driven image editing, a key area in computer graphics, faces unique challenges. A major challenge is making simultaneous edits across multiple objects or attributes. Applying these methods sequentially for multi-attribute edits increases computational demands and efficiency losses. In this paper, we address these challenges with significant contributions. Our main contribution is the development of ParallelEdits, a method that seamlessly manages simultaneous edits across multiple attributes. In contrast to previous approaches, ParallelEdits not only preserves the quality of single attribute edits but also significantly improves the performance of multitasking edits. This is achieved through innovative attention distribution mechanism and multi-branch design that operates across several processing heads. Additionally, we introduce the PIE-Bench++ dataset, an expansion of the original PIE-Bench dataset, to better support evaluating image-editing tasks involving multiple objects and attributes simultaneously. This dataset is a benchmark for evaluating text-driven image editing methods in multifaceted scenarios.

2406.01163

Bhavya Sukhija

Lenart Treven, Bhavya Sukhija, Yarden As, Florian D\"orfler, Andreas Krause

When to Sense and Control? A Time-adaptive Approach for Continuous-Time RL

cs.LG

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Reinforcement learning (RL) excels in optimizing policies for discrete-time Markov decision processes (MDP). However, various systems are inherently continuous in time, making discrete-time MDPs an inexact modeling choice. In many applications, such as greenhouse control or medical treatments, each interaction (measurement or switching of action) involves manual intervention and thus is inherently costly. Therefore, we generally prefer a time-adaptive approach with fewer interactions with the system. In this work, we formalize an RL framework, Time-adaptive Control & Sensing (TaCoS), that tackles this challenge by optimizing over policies that besides control predict the duration of its application. Our formulation results in an extended MDP that any standard RL algorithm can solve. We demonstrate that state-of-the-art RL algorithms trained on TaCoS drastically reduce the interaction amount over their discrete-time counterpart while retaining the same or improved performance, and exhibiting robustness over discretization frequency. Finally, we propose OTaCoS, an efficient model-based algorithm for our setting. We show that OTaCoS enjoys sublinear regret for systems with sufficiently smooth dynamics and empirically results in further sample-efficiency gains.

2406.01175

Bhavya Sukhija

Bhavya Sukhija, Lenart Treven, Florian D\"orfler, Stelian Coros, Andreas Krause

NeoRL: Efficient Exploration for Nonepisodic RL

cs.LG

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

We study the problem of nonepisodic reinforcement learning (RL) for nonlinear dynamical systems, where the system dynamics are unknown and the RL agent has to learn from a single trajectory, i.e., without resets. We propose Nonepisodic Optimistic RL (NeoRL), an approach based on the principle of optimism in the face of uncertainty. NeoRL uses well-calibrated probabilistic models and plans optimistically w.r.t. the epistemic uncertainty about the unknown dynamics. Under continuity and bounded energy assumptions on the system, we provide a first-of-its-kind regret bound of $\setO(\beta_T \sqrt{T \Gamma_T})$ for general nonlinear systems with Gaussian process dynamics. We compare NeoRL to other baselines on several deep RL environments and empirically demonstrate that NeoRL achieves the optimal average cost while incurring the least regret.

2406.01257

Kairan Zhao

Kairan Zhao, Meghdad Kurmanji, George-Octavian B\u{a}rbulescu, Eleni Triantafillou, Peter Triantafillou

What makes unlearning hard and what to do about it

Accepted at NeurIPS 2024

cs.LG

http://creativecommons.org/licenses/by/4.0/

Machine unlearning is the problem of removing the effect of a subset of training data (the ''forget set'') from a trained model without damaging the model's utility e.g. to comply with users' requests to delete their data, or remove mislabeled, poisoned or otherwise problematic data. With unlearning research still being at its infancy, many fundamental open questions exist: Are there interpretable characteristics of forget sets that substantially affect the difficulty of the problem? How do these characteristics affect different state-of-the-art algorithms? With this paper, we present the first investigation aiming to answer these questions. We identify two key factors affecting unlearning difficulty and the performance of unlearning algorithms. Evaluation on forget sets that isolate these identified factors reveals previously-unknown behaviours of state-of-the-art algorithms that don't materialize on random forget sets. Based on our insights, we develop a framework coined Refined-Unlearning Meta-algorithm (RUM) that encompasses: (i) refining the forget set into homogenized subsets, according to different characteristics; and (ii) a meta-algorithm that employs existing algorithms to unlearn each subset and finally delivers a model that has unlearned the overall forget set. We find that RUM substantially improves top-performing unlearning algorithms. Overall, we view our work as an important step in (i) deepening our scientific understanding of unlearning and (ii) revealing new pathways to improving the state-of-the-art.

2406.01563

Fangcong Yin

Fangcong Yin, Xi Ye, Greg Durrett

LoFiT: Localized Fine-tuning on LLM Representations

NeurIPS 2024 Camera Ready

cs.CL

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Recent work in interpretability shows that large language models (LLMs) can be adapted for new tasks in a learning-free way: it is possible to intervene on LLM representations to elicit desired behaviors for alignment. For instance, adding certain bias vectors to the outputs of certain attention heads is reported to boost the truthfulness of models. In this work, we show that localized fine-tuning serves as an effective alternative to such representation intervention methods. We introduce a framework called Localized Fine-Tuning on LLM Representations (LoFiT), which identifies a subset of attention heads that are most important for learning a specific task, then trains offset vectors to add to the model's hidden representations at those selected heads. LoFiT localizes to a sparse set of heads (3%-10%) and learns the offset vectors from limited training data, comparable to the settings used for representation intervention. For truthfulness and reasoning tasks, we find that LoFiT's intervention vectors are more effective for LLM adaptation than vectors from representation intervention methods such as Inference-time Intervention. We also find that the localization step is important: selecting a task-specific set of attention heads can lead to higher performance than intervening on heads selected for a different task. Finally, across 7 tasks we study, LoFiT achieves comparable performance to other parameter-efficient fine-tuning methods such as LoRA, despite modifying 20x-200x fewer parameters than these methods.

2406.02389

Uddalok Sen

Maaike Rump, Christian Diddens, Uddalok Sen, Michel Versluis, Detlef Lohse, and Tim Segers

Role of surfactants on droplet formation in piezoacoustic inkjet printing across microsecond-to-second timescales

13 pages, 8 figures

physics.flu-dyn cond-mat.soft

http://creativecommons.org/licenses/by/4.0/

In piezo acoustic drop-on-demand inkjet printing a single droplet is produced for each piezo driving pulse. This droplet is typically multicomponent, including surfactants to control the spreading and drying of the droplet on the substrate. However, the role of these surfactants on the droplet formation process remains rather elusive. Surfactant concentration gradients may manifest across microsecond-to-second timescales, spanning both the rapid ejection of ink from the nozzle exit and the comparatively slower idling timescale governing the firing of successive droplets. In the present work, we study the influence of surfactants on droplet formation across 6 orders of magnitude in time. To this end, we visualize the microsecond droplet formation process using stroboscopic laser-induced fluorescence microscopy while we vary the nozzle idle time. Our results show that increasing the idle time up to O(1) s affects only the break-up dynamics of the inkjet but not its velocity. By contrast, for idle times $>$ O(1) s, both the break-up dynamics are altered and the velocity of the inkjet increases. We show that the increased velocity results from a decreased surface tension of the ejected droplet, which we extracted from the observed shape oscillations of the jetted droplets in flight. The measured decrease in surface tension is surprising as the $\mu$s timescale of droplet formation is much faster than the typical ms-to-s timescale of surfactant adsorption. By varying the bulk surfactant concentration, we show that the fast decrease in surface tension results from a local surfactant concentration increase to more than 200 times the CMC. Our results suggest that a local high concentration of surfactant allows for surfactant adsorption to the interface of an inkjet at the $\mu$s-to-ms timescale, which is much faster than the typical ms-to-s timescale of surfactant adsorption.

2406.02427

Kenneth Goodenough

Kenneth Goodenough, Aqil Sajjad, Eneet Kaur, Saikat Guha, Don Towsley

Bipartite entanglement of noisy stabilizer states through the lens of stabilizer codes

8 pages, 4 figures

quant-ph

http://creativecommons.org/licenses/by/4.0/

Stabilizer states are a prime resource for a number of applications in quantum information science, such as secret-sharing and measurement-based quantum computation. This motivates us to study the entanglement of noisy stabilizer states across a bipartition. We show that the spectra of the corresponding reduced states can be expressed in terms of properties of an associated stabilizer code. In particular, this allows us to show that the coherent information is related to the so-called syndrome entropy of the underlying code. We use this viewpoint to find stabilizer states that are resilient against noise, allowing for more robust entanglement distribution in near-term quantum networks. We specialize our results to the case of graph states, where the found connections with stabilizer codes reduces back to classical linear codes for dephasing noise. On our way we provide an alternative proof of the fact that every qubit stabilizer code is equivalent up to single-qubit Clifford gates to a graph code.

2406.02460

Martin Spitz

Sebastian Herr and Mihaela Ifrim and Martin Spitz

Modified scattering for the three dimensional Maxwell-Dirac system

64 pages; v2: minor corrections; v3: processing errors concerning the tikz pictures fixed

math.AP math.CA

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

In this work we prove global well-posedness for the massive Maxwell-Dirac system in the Lorenz gauge in $\mathbb{R}^{1+3}$, for small, sufficiently smooth and decaying initial data, as well as modified scattering for the solutions. Heuristically we exploit the close connection between the massive Maxwell-Dirac and the wave-Klein-Gordon equations, while developing a novel approach which applies directly at the level of the Dirac equations. The modified scattering result follows from a precise description of the asymptotic behavior of the solutions inside the light cone, which we derive via the method of testing with wave packets of Ifrim-Tataru.

2406.02462

Jason Hu

Jason Hu, Bowen Song, Xiaojian Xu, Liyue Shen, Jeffrey A. Fessler

Learning Image Priors through Patch-based Diffusion Models for Solving Inverse Problems

cs.CV cs.AI

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Diffusion models can learn strong image priors from underlying data distribution and use them to solve inverse problems, but the training process is computationally expensive and requires lots of data. Such bottlenecks prevent most existing works from being feasible for high-dimensional and high-resolution data such as 3D images. This paper proposes a method to learn an efficient data prior for the entire image by training diffusion models only on patches of images. Specifically, we propose a patch-based position-aware diffusion inverse solver, called PaDIS, where we obtain the score function of the whole image through scores of patches and their positional encoding and utilize this as the prior for solving inverse problems. First of all, we show that this diffusion model achieves an improved memory efficiency and data efficiency while still maintaining the capability to generate entire images via positional encoding. Additionally, the proposed PaDIS model is highly flexible and can be plugged in with different diffusion inverse solvers (DIS). We demonstrate that the proposed PaDIS approach enables solving various inverse problems in both natural and medical image domains, including CT reconstruction, deblurring, and superresolution, given only patch-based priors. Notably, PaDIS outperforms previous DIS methods trained on entire image priors in the case of limited training data, demonstrating the data efficiency of our proposed approach by learning patch-based prior.

2406.03214

Dorothee Mader

Doroth\'ee S. Mader, Richarda Niemann, Martin Wolf, Sebastian F. Maehrlein and Alexander Paarmann

Sum-Frequency Generation Spectro-Microscopy in the Reststrahlen Band of Wurtzite-type Aluminum Nitride

6 pages, 3 figures, SI upon request to the authors

10.1063/5.0220619

physics.optics physics.ins-det

http://creativecommons.org/licenses/by-nc-sa/4.0/

Nonlinear-optical microscopy and spectroscopy provide detailed spatial and spectroscopic contrast, specifically sensitive to structural symmetry and order. Ferroics, in particular, have been widely studied using second harmonic generation imaging, which provides detailed information on domain structures but typically lacks spectroscopic detail. In contrast, infrared-visible sum-frequency generation (SFG) spectroscopy reveals details of the atomic structure and bonding via vibrational resonances, but conventionally lacks spatial information. In this work, we combine the benefits of nonlinear optical imaging and SFG spectroscopy by employing SFG spectro-microscopy using an infrared free-electron laser. Specifically, we demonstrate the feasibility of SFG spectro-microscopy for spectroscopy using in-plane anisotropic wurtzite-type aluminum nitride as a model system. We find the experimental spectra to agree well with our theoretical calculations and we show the potential of our microscope to provide spatially resolved spectroscopic information in inhomogeneous systems such as ferroics and their domains in the near future.

2406.03371

Mohsen Fathi

Uktamjon Uktamov, Mohsen Fathi, Javlon Rayimbaev and Ahmadjon Abdujabbarov

Orbits of particles with magnetic dipole moment around magnetized Schwarzschild black holes: Applications to S2 star orbit

17 pages, 10 figures, 6 tables

Phys. Rev. D 110, 084084 (2024)

10.1103/PhysRevD.110.084084

gr-qc astro-ph.HE hep-th

http://creativecommons.org/licenses/by/4.0/

This study provides a comprehensive analytical investigation of the bound and unbound motion of magnetized particles orbiting a Schwarzschild black hole immersed in an external asymptotically uniform magnetic field, which includes all conceivable types of bounded and unbounded orbits. In particular, for planetary orbits, we perform a comparative analysis of our findings with the observed position of the S2 star carrying magnetic dipole moment around Sagittarius A* (Sgr A*). We found maximum and minimum values for the parameter of magnetic interaction between the magnetic dipole of the star and the external magnetic field, as well as the energy and angular momentum of the S2 star. As a result, we obtain estimations of the magnetic dipole of the star in order of $10^6 \rm \ G\cdot cm^{3}$. Additionally, we explore deflecting trajectories akin to gravitational Rutherford scattering. In obtaining the solutions for the orbital equations, we articulate the elliptic integrals and Jacobi elliptic functions, and our study is augmented by illustrative figures and simulations.

2406.03636

Federico Mora

Federico Mora, Justin Wong, Haley Lepe, Sahil Bhatia, Karim Elmaaroufi, George Varghese, Joseph E. Gonzalez, Elizabeth Polgreen, Sanjit A. Seshia

Synthetic Programming Elicitation for Text-to-Code in Very Low-Resource Programming and Formal Languages

14 pages, 6 figures, 1 table

cs.PL cs.LG

http://creativecommons.org/licenses/by/4.0/

Recent advances in large language models (LLMs) for code applications have demonstrated remarkable zero-shot fluency and instruction following on challenging code related tasks ranging from test case generation to self-repair. Unsurprisingly, however, models struggle to compose syntactically valid programs in programming languages unrepresented in pre-training, referred to as very low-resource Programming Languages (VLPLs). VLPLs appear in crucial settings, including domain-specific languages for internal tools, tool-chains for legacy languages, and formal verification frameworks. Inspired by a technique called natural programming elicitation, we propose designing an intermediate language that LLMs "naturally" know how to use and which can be automatically compiled to a target VLPL. When LLMs generate code that lies outside of this intermediate language, we use compiler techniques to repair the code into programs in the intermediate language. Overall, we introduce \emph{synthetic programming elicitation and compilation} (SPEAC), an approach that enables LLMs to generate syntactically valid code even for VLPLs. We empirically evaluate the performance of SPEAC in a case study for the UCLID5 formal verification language and find that, compared to existing retrieval and fine-tuning baselines, SPEAC produces syntactically correct programs more frequently and without sacrificing semantic correctness.

2406.04230

Matt Allen

Matthew J Allen, Francisco Dorr, Joseph Alejandro Gallego Mejia, Laura Mart\'inez-Ferrer, Anna Jungbluth, Freddie Kalaitzis, Ra\'ul Ramos-Poll\'an

M3LEO: A Multi-Modal, Multi-Label Earth Observation Dataset Integrating Interferometric SAR and Multispectral Data

10 pages, 5 figures

cs.CV cs.AI

http://creativecommons.org/licenses/by-sa/4.0/

Satellite-based remote sensing has revolutionised the way we address global challenges. Huge quantities of Earth Observation (EO) data are generated by satellite sensors daily, but processing these large datasets for use in ML pipelines is technically and computationally challenging. While some preprocessed Earth observation datasets exist, their content is often limited to optical or near-optical wavelength data, which is ineffective at night or in adverse weather conditions. Synthetic Aperture Radar (SAR), an active sensing technique based on microwave length radiation, offers a viable alternative. However, the application of machine learning to SAR has been limited due to a lack of ML-ready data and pipelines, particularly for the full diversity of SAR data, including polarimetry, coherence and interferometry. In this work, we introduce M3LEO, a multi-modal, multi-label Earth observation dataset that includes polarimetric, interferometric, and coherence SAR data derived from Sentinel-1, alongside multispectral Sentinel-2 imagery and auxiliary data describing terrain properties such as land use. M3LEO spans approximately 17M 4x4 km data chips from six diverse geographic regions. The dataset is complemented by a flexible PyTorch Lightning framework configured using Hydra to accommodate its use across diverse ML applications in Earth observation. We provide tools to process any dataset available on popular platforms such as Google Earth Engine for seamless integration with our framework. We show that the distribution shift in self-supervised embeddings is substantial across geographic regions, even when controlling for terrain properties. Data: huggingface.co/M3LEO, Code: github.com/spaceml-org/M3LEO.

2406.04312

Shyamgopal Karthik

Luca Eyring, Shyamgopal Karthik, Karsten Roth, Alexey Dosovitskiy, Zeynep Akata

ReNO: Enhancing One-step Text-to-Image Models through Reward-based Noise Optimization

NeurIPS 2024

cs.CV

http://creativecommons.org/licenses/by/4.0/

Text-to-Image (T2I) models have made significant advancements in recent years, but they still struggle to accurately capture intricate details specified in complex compositional prompts. While fine-tuning T2I models with reward objectives has shown promise, it suffers from "reward hacking" and may not generalize well to unseen prompt distributions. In this work, we propose Reward-based Noise Optimization (ReNO), a novel approach that enhances T2I models at inference by optimizing the initial noise based on the signal from one or multiple human preference reward models. Remarkably, solving this optimization problem with gradient ascent for 50 iterations yields impressive results on four different one-step models across two competitive benchmarks, T2I-CompBench and GenEval. Within a computational budget of 20-50 seconds, ReNO-enhanced one-step models consistently surpass the performance of all current open-source Text-to-Image models. Extensive user studies demonstrate that our model is preferred nearly twice as often compared to the popular SDXL model and is on par with the proprietary Stable Diffusion 3 with 8B parameters. Moreover, given the same computational resources, a ReNO-optimized one-step model outperforms widely-used open-source models such as SDXL and PixArt-$\alpha$, highlighting the efficiency and effectiveness of ReNO in enhancing T2I model performance at inference time. Code is available at https://github.com/ExplainableML/ReNO.

2406.04435

Benjamin Wild

Benjamin W. Wild and Roderick Edwards

Entropy bounds for Glass networks

36 pages, 14 figures

math.DS

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

We propose that chaotic Glass networks (a class of piecewise-linear Ordinary Differential Equations) are good candidates for the design of true random number generators. A Glass network design has the advantage of involving only standard Boolean logic gates. Furthermore, an already chaotic (deterministic) system combined with random ``jitter'' due to thermal noise can be used to generate random bit sequences in a more robust way than noisy limit-cycle oscillators. Since the goal is to generate bit sequences with as large a positive entropy as possible, it is desirable to have a theoretical method to assess the irregularity of a large class of networks. We develop a procedure here to calculate good upper bounds on the entropy of a Glass network, by means of symbolic representations of the continuous dynamics. Our method improves on a result by Farcot (2006), and allows in principle for an arbitrary level of precision by refinements of the estimate, and we show that in the limiting case, these estimates converge to the true entropy of the symbolic system corresponding to the continuous dynamics. As a check on the method, we demonstrate for an example network that our upper bound after only a few refinement steps is very close to the entropy estimated from a long numerical simulation.

2406.04823

David Samuel

David Samuel

BERTs are Generative In-Context Learners

26 pages, NeurIPS 2024

cs.CL cs.AI

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

While in-context learning is commonly associated with causal language models, such as GPT, we demonstrate that this capability also 'emerges' in masked language models. Through an embarrassingly simple inference technique, we enable an existing masked model, DeBERTa, to perform generative tasks without additional training or architectural changes. Our evaluation reveals that the masked and causal language models behave very differently, as they clearly outperform each other on different categories of tasks. These complementary strengths suggest that the field's focus on causal models for in-context learning may be limiting - both architectures can develop these capabilities, but with distinct advantages; pointing toward promising hybrid approaches that combine the strengths of both objectives.

2406.05061

Parnian Kassraie

Parnian Kassraie, Aram-Alexandre Pooladian, Michal Klein, James Thornton, Jonathan Niles-Weed, Marco Cuturi

Progressive Entropic Optimal Transport Solvers

In proceedings of the 38th conference on Advances in Neural Information Processing Systems (NeurIPs), 23 pages, 14 figures

stat.ML cs.LG

http://creativecommons.org/licenses/by-nc-sa/4.0/

Optimal transport (OT) has profoundly impacted machine learning by providing theoretical and computational tools to realign datasets. In this context, given two large point clouds of sizes $n$ and $m$ in $\mathbb{R}^d$, entropic OT (EOT) solvers have emerged as the most reliable tool to either solve the Kantorovich problem and output a $n\times m$ coupling matrix, or to solve the Monge problem and learn a vector-valued push-forward map. While the robustness of EOT couplings/maps makes them a go-to choice in practical applications, EOT solvers remain difficult to tune because of a small but influential set of hyperparameters, notably the omnipresent entropic regularization strength $\varepsilon$. Setting $\varepsilon$ can be difficult, as it simultaneously impacts various performance metrics, such as compute speed, statistical performance, generalization, and bias. In this work, we propose a new class of EOT solvers (ProgOT), that can estimate both plans and transport maps. We take advantage of several opportunities to optimize the computation of EOT solutions by dividing mass displacement using a time discretization, borrowing inspiration from dynamic OT formulations, and conquering each of these steps using EOT with properly scheduled parameters. We provide experimental evidence demonstrating that ProgOT is a faster and more robust alternative to standard solvers when computing couplings at large scales, even outperforming neural network-based approaches. We also prove statistical consistency of our approach for estimating optimal transport maps.

2406.05175

Victor Yon

Victor Yon, Bastien Galaup, Claude Rohrbacher, Joffrey Rivard, Cl\'ement Godfrin, Ruoyu Li, Stefan Kubicek, Kristiaan De Greve, Louis Gaudreau, Eva Dupont-Ferrier, Yann Beilliard, Roger G. Melko, Dominique Drouin

Robust quantum dots charge autotuning using neural network uncertainty

12 pages (main) + 14 pages (supplementary)

10.1088/2632-2153/ad88d5

quant-ph cs.LG

http://creativecommons.org/licenses/by/4.0/

This study presents a machine-learning-based procedure to automate the charge tuning of semiconductor spin qubits with minimal human intervention, addressing one of the significant challenges in scaling up quantum dot technologies. This method exploits artificial neural networks to identify noisy transition lines in stability diagrams, guiding a robust exploration strategy leveraging neural networks' uncertainty estimations. Tested across three distinct offline experimental datasets representing different single quantum dot technologies, the approach achieves over 99% tuning success rate in optimal cases, where more than 10% of the success is directly attributable to uncertainty exploitation. The challenging constraints of small training sets containing high diagram-to-diagram variability allowed us to evaluate the capabilities and limits of the proposed procedure.

2406.05723

Zheng Chen

Zheng Chen, Haotong Qin, Yong Guo, Xiongfei Su, Xin Yuan, Linghe Kong, Yulun Zhang

Binarized Diffusion Model for Image Super-Resolution

Accepted to NeurIPS 2024. Code is available at https://github.com/zhengchen1999/BI-DiffSR

cs.CV

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Advanced diffusion models (DMs) perform impressively in image super-resolution (SR), but the high memory and computational costs hinder their deployment. Binarization, an ultra-compression algorithm, offers the potential for effectively accelerating DMs. Nonetheless, due to the model structure and the multi-step iterative attribute of DMs, existing binarization methods result in significant performance degradation. In this paper, we introduce a novel binarized diffusion model, BI-DiffSR, for image SR. First, for the model structure, we design a UNet architecture optimized for binarization. We propose the consistent-pixel-downsample (CP-Down) and consistent-pixel-upsample (CP-Up) to maintain dimension consistent and facilitate the full-precision information transfer. Meanwhile, we design the channel-shuffle-fusion (CS-Fusion) to enhance feature fusion in skip connection. Second, for the activation difference across timestep, we design the timestep-aware redistribution (TaR) and activation function (TaA). The TaR and TaA dynamically adjust the distribution of activations based on different timesteps, improving the flexibility and representation alability of the binarized module. Comprehensive experiments demonstrate that our BI-DiffSR outperforms existing binarization methods. Code is released at: https://github.com/zhengchen1999/BI-DiffSR.

2406.05768

Qingsong Xie

Qingsong Xie, Zhenyi Liao, Zhijie Deng, Chen chen, Haonan Lu

TLCM: Training-efficient Latent Consistency Model for Image Generation with 2-8 Steps

cs.CV cs.AI

http://creativecommons.org/licenses/by/4.0/

Distilling latent diffusion models (LDMs) into ones that are fast to sample from is attracting growing research interest. However, the majority of existing methods face two critical challenges: (1) They hinge on long training using a huge volume of real data. (2) They routinely lead to quality degradation for generation, especially in text-image alignment. This paper proposes a novel training-efficient Latent Consistency Model (TLCM) to overcome these challenges. Our method first accelerates LDMs via data-free multistep latent consistency distillation (MLCD), and then data-free latent consistency distillation is proposed to efficiently guarantee the inter-segment consistency in MLCD. Furthermore, we introduce bags of techniques, e.g., distribution matching, adversarial learning, and preference learning, to enhance TLCM's performance at few-step inference without any real data. TLCM demonstrates a high level of flexibility by enabling adjustment of sampling steps within the range of 2 to 8 while still producing competitive outputs compared to full-step approaches. Notably, TLCM enjoys the data-free merit by employing synthetic data from the teacher for distillation. With just 70 training hours on an A100 GPU, a 3-step TLCM distilled from SDXL achieves an impressive CLIP Score of 33.68 and an Aesthetic Score of 5.97 on the MSCOCO-2017 5K benchmark, surpassing various accelerated models and even outperforming the teacher model in human preference metrics. We also demonstrate the versatility of TLCMs in applications including image style transfer, controllable generation, and Chinese-to-image generation.

2406.05983

Uihyeop Shin

Ui-Hyeop Shin, Sangyoun Lee, Taehan Kim, Hyung-Min Park

Separate and Reconstruct: Asymmetric Encoder-Decoder for Speech Separation

Accepted to NeurIPS 2024; Project Page: https://dmlguq456.github.io/SepReformer_Demo

eess.AS

http://creativecommons.org/licenses/by/4.0/

In speech separation, time-domain approaches have successfully replaced the time-frequency domain with latent sequence feature from a learnable encoder. Conventionally, the feature is separated into speaker-specific ones at the final stage of the network. Instead, we propose a more intuitive strategy that separates features earlier by expanding the feature sequence to the number of speakers as an extra dimension. To achieve this, an asymmetric strategy is presented in which the encoder and decoder are partitioned to perform distinct processing in separation tasks. The encoder analyzes features, and the output of the encoder is split into the number of speakers to be separated. The separated sequences are then reconstructed by the weight-shared decoder, which also performs cross-speaker processing. Without relying on speaker information, the weight-shared network in the decoder directly learns to discriminate features using a separation objective. In addition, to improve performance, traditional methods have extended the sequence length, leading to the adoption of dual-path models, which handle the much longer sequence effectively by segmenting it into chunks. To address this, we introduce global and local Transformer blocks that can directly handle long sequences more efficiently without chunking and dual-path processing. The experimental results demonstrated that this asymmetric structure is effective and that the combination of proposed global and local Transformer can sufficiently replace the role of inter- and intra-chunk processing in dual-path structure. Finally, the presented model combining both of these achieved state-of-the-art performance with much less computation in various benchmark datasets.

2406.06196

Andrew Bean

Andrew M. Bean, Simi Hellsten, Harry Mayne, Jabez Magomere, Ethan A. Chi, Ryan Chi, Scott A. Hale, Hannah Rose Kirk

LINGOLY: A Benchmark of Olympiad-Level Linguistic Reasoning Puzzles in Low-Resource and Extinct Languages

Oral presentation at NeurIPS 2024 Datasets and Benchmarks Track. 10 pages, 5 figures, 22 pages supplemental materials

cs.CL

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

In this paper, we present the LingOly benchmark, a novel benchmark for advanced reasoning abilities in large language models. Using challenging Linguistic Olympiad puzzles, we evaluate (i) capabilities for in-context identification and generalisation of linguistic patterns in very low-resource or extinct languages, and (ii) abilities to follow complex task instructions. The LingOly benchmark covers more than 90 mostly low-resource languages, minimising issues of data contamination, and contains 1,133 problems across 6 formats and 5 levels of human difficulty. We assess performance with both direct accuracy and comparison to a no-context baseline to penalise memorisation. Scores from 11 state-of-the-art LLMs demonstrate the benchmark to be challenging, and models perform poorly on the higher difficulty problems. On harder problems, even the top model only achieved 38.7% accuracy, a 24.7% improvement over the no-context baseline. Large closed models typically outperform open models, and in general, the higher resource the language, the better the scores. These results indicate, in absence of memorisation, true multi-step out-of-domain reasoning remains a challenge for current language models.

2406.06246

Alaia Solko-Breslin

Alaia Solko-Breslin, Seewon Choi, Ziyang Li, Neelay Velingker, Rajeev Alur, Mayur Naik, Eric Wong

Data-Efficient Learning with Neural Programs

cs.LG

http://creativecommons.org/licenses/by/4.0/

Many computational tasks can be naturally expressed as a composition of a DNN followed by a program written in a traditional programming language or an API call to an LLM. We call such composites "neural programs" and focus on the problem of learning the DNN parameters when the training data consist of end-to-end input-output labels for the composite. When the program is written in a differentiable logic programming language, techniques from neurosymbolic learning are applicable, but in general, the learning for neural programs requires estimating the gradients of black-box components. We present an algorithm for learning neural programs, called ISED, that only relies on input-output samples of black-box components. For evaluation, we introduce new benchmarks that involve calls to modern LLMs such as GPT-4 and also consider benchmarks from the neurosymbolic learning literature. Our evaluation shows that for the latter benchmarks, ISED has comparable performance to state-of-the-art neurosymbolic frameworks. For the former, we use adaptations of prior work on gradient approximations of black-box components as a baseline, and show that ISED achieves comparable accuracy but in a more data- and sample-efficient manner.

2406.06714

Vivek Myers

Michelle Pan, Mariah Schrum, Vivek Myers, Erdem B{\i}y{\i}k, Anca Dragan

Coprocessor Actor Critic: A Model-Based Reinforcement Learning Approach For Adaptive Brain Stimulation

Proceedings of the 41st International Conference on Machine Learning (ICML 2024)

International Conference on Machine Learning 2024

cs.LG cs.AI cs.HC

http://creativecommons.org/licenses/by/4.0/

Adaptive brain stimulation can treat neurological conditions such as Parkinson's disease and post-stroke motor deficits by influencing abnormal neural activity. Because of patient heterogeneity, each patient requires a unique stimulation policy to achieve optimal neural responses. Model-free reinforcement learning (MFRL) holds promise in learning effective policies for a variety of similar control tasks, but is limited in domains like brain stimulation by a need for numerous costly environment interactions. In this work we introduce Coprocessor Actor Critic, a novel, model-based reinforcement learning (MBRL) approach for learning neural coprocessor policies for brain stimulation. Our key insight is that coprocessor policy learning is a combination of learning how to act optimally in the world and learning how to induce optimal actions in the world through stimulation of an injured brain. We show that our approach overcomes the limitations of traditional MFRL methods in terms of sample efficiency and task success and outperforms baseline MBRL approaches in a neurologically realistic model of an injured brain.

2406.06818

Yuanjie Shi

Yuanjie Shi, Subhankar Ghosh, Taha Belkhouja, Janardhan Rao Doppa and Yan Yan

Conformal Prediction for Class-wise Coverage via Augmented Label Rank Calibration

cs.LG

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Conformal prediction (CP) is an emerging uncertainty quantification framework that allows us to construct a prediction set to cover the true label with a pre-specified marginal or conditional probability. Although the valid coverage guarantee has been extensively studied for classification problems, CP often produces large prediction sets which may not be practically useful. This issue is exacerbated for the setting of class-conditional coverage on imbalanced classification tasks with many and/or imbalanced classes. This paper proposes the Rank Calibrated Class-conditional CP (RC3P) algorithm to reduce the prediction set sizes to achieve class-conditional coverage, where the valid coverage holds for each class. In contrast to the standard class-conditional CP (CCP) method that uniformly thresholds the class-wise conformity score for each class, the augmented label rank calibration step allows RC3P to selectively iterate this class-wise thresholding subroutine only for a subset of classes whose class-wise top-k error is small. We prove that agnostic to the classifier and data distribution, RC3P achieves class-wise coverage. We also show that RC3P reduces the size of prediction sets compared to the CCP method. Comprehensive experiments on multiple real-world datasets demonstrate that RC3P achieves class-wise coverage and 26.25% reduction in prediction set sizes on average.

2406.06925

Wenchuan Yang

Wenchuan Yang, Cheng Yang, Jichao Li, Yuejin Tan, Xin Lu, Chuan Shi

Non-autoregressive Personalized Bundle Generation

Submitted to Information Processing & Management

10.1016/j.ipm.2024.103814

cs.LG cs.IR

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

The personalized bundle generation problem, which aims to create a preferred bundle for user from numerous candidate items, receives increasing attention in recommendation. However, existing works ignore the order-invariant nature of the bundle and adopt sequential modeling methods as the solution, which might introduce inductive bias and cause a large latency in prediction. To address this problem, we propose to perform the bundle generation via non-autoregressive mechanism and design a novel encoder-decoder framework named BundleNAT, which can effectively output the targeted bundle in one-shot without relying on any inherent order. In detail, instead of learning sequential dependency, we propose to adopt pre-training techniques and graph neural network to fully embed user-based preference and item-based compatibility information, and use a self-attention based encoder to further extract global dependency pattern. We then design a permutation-equivariant decoding architecture that is able to directly output the desired bundle in a one-shot manner. Experiments on three real-world datasets from Youshu and Netease show the proposed BundleNAT significantly outperforms the current state-of-the-art methods in average by up to 35.92%, 10.97% and 23.67% absolute improvements in Precision, Precision+, and Recall, respectively.

2406.07085

Zhongzhen Huang

Zhongzhen Huang, Yankai Jiang, Rongzhao Zhang, Shaoting Zhang, Xiaofan Zhang

CAT: Coordinating Anatomical-Textual Prompts for Multi-Organ and Tumor Segmentation

cs.CV

http://creativecommons.org/licenses/by/4.0/

Existing promptable segmentation methods in the medical imaging field primarily consider either textual or visual prompts to segment relevant objects, yet they often fall short when addressing anomalies in medical images, like tumors, which may vary greatly in shape, size, and appearance. Recognizing the complexity of medical scenarios and the limitations of textual or visual prompts, we propose a novel dual-prompt schema that leverages the complementary strengths of visual and textual prompts for segmenting various organs and tumors. Specifically, we introduce CAT, an innovative model that Coordinates Anatomical prompts derived from 3D cropped images with Textual prompts enriched by medical domain knowledge. The model architecture adopts a general query-based design, where prompt queries facilitate segmentation queries for mask prediction. To synergize two types of prompts within a unified framework, we implement a ShareRefiner, which refines both segmentation and prompt queries while disentangling the two types of prompts. Trained on a consortium of 10 public CT datasets, CAT demonstrates superior performance in multiple segmentation tasks. Further validation on a specialized in-house dataset reveals the remarkable capacity of segmenting tumors across multiple cancer stages. This approach confirms that coordinating multimodal prompts is a promising avenue for addressing complex scenarios in the medical domain.

2406.07451

Xiaoyan Hu

Xiaoyan Hu, Ho-fung Leung, and Farzan Farnia

An Optimism-based Approach to Online Evaluation of Generative Models

arXiv version

cs.LG

http://creativecommons.org/licenses/by/4.0/

Existing frameworks for evaluating and comparing generative models typically target an offline setting, where the evaluator has access to full batches of data produced by the models. However, in many practical scenarios, the goal is to identify the best model using the fewest generated samples to minimize the costs of querying data from the models. Such an online comparison is challenging with current offline assessment methods. In this work, we propose an online evaluation framework to find the generative model that maximizes a standard assessment score among a group of available models. Our method uses an optimism-based multi-armed bandit framework to identify the model producing data with the highest evaluation score, quantifying the quality and diversity of generated data. Specifically, we study the online assessment of generative models based on the Fr\'echet Inception Distance (FID) and Inception Score (IS) metrics and propose the FID-UCB and IS-UCB algorithms leveraging the upper confidence bound approach in online learning. We prove sub-linear regret bounds for these algorithms and present numerical results on standard image datasets, demonstrating their effectiveness in identifying the score-maximizing generative model.

2406.07592

Oliver Eberle Dr

Farnoush Rezaei Jafari, Gr\'egoire Montavon, Klaus-Robert M\"uller, Oliver Eberle

MambaLRP: Explaining Selective State Space Sequence Models

cs.LG cs.AI stat.ML

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Recent sequence modeling approaches using selective state space sequence models, referred to as Mamba models, have seen a surge of interest. These models allow efficient processing of long sequences in linear time and are rapidly being adopted in a wide range of applications such as language modeling, demonstrating promising performance. To foster their reliable use in real-world scenarios, it is crucial to augment their transparency. Our work bridges this critical gap by bringing explainability, particularly Layer-wise Relevance Propagation (LRP), to the Mamba architecture. Guided by the axiom of relevance conservation, we identify specific components in the Mamba architecture, which cause unfaithful explanations. To remedy this issue, we propose MambaLRP, a novel algorithm within the LRP framework, which ensures a more stable and reliable relevance propagation through these components. Our proposed method is theoretically sound and excels in achieving state-of-the-art explanation performance across a diverse range of models and datasets. Moreover, MambaLRP facilitates a deeper inspection of Mamba architectures, uncovering various biases and evaluating their significance. It also enables the analysis of previous speculations regarding the long-range capabilities of Mamba models.

2406.07608

Chenyuan Li

Chenyuan Li, Davide Valentinis, Aavishkar A. Patel, Haoyu Guo, J\"org Schmalian, Subir Sachdev and Ilya Esterlis

Strange metal and superconductor in the two-dimensional Yukawa-Sachdev-Ye-Kitaev model

7+18 pages, 5+12 figures; references added

Physical Review Letters 133, 186502 (2024)

10.1103/PhysRevLett.133.186502

cond-mat.str-el cond-mat.supr-con hep-th

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

The two-dimensional Yukawa-Sachdev-Ye-Kitaev (2d-YSYK) model provides a universal theory of quantum phase transitions in metals in the presence of quenched random spatial fluctuations in the local position of the quantum critical point. It has a Fermi surface coupled to a scalar field by spatially random Yukawa interactions. We present full numerical solutions of a self-consistent disorder averaged analysis of the 2d-YSYK model in both the normal and superconducting states, obtaining electronic spectral functions, frequency-dependent conductivity, and superfluid stiffness. Our results reproduce key aspects of observations in the cuprates as analyzed by Michon et al. (arXiv:2205.04030). We also find a regime of increasing zero temperature superfluid stiffness with decreasing superconducting critical temperature, as is observed in bulk cuprates.

2406.07933

Chris Yuhao Liu

Chris Yuhao Liu, Yaxuan Wang, Jeffrey Flanigan, Yang Liu

Large Language Model Unlearning via Embedding-Corrupted Prompts

NeurIPS 2024 Poster

cs.CL cs.AI cs.LG

http://creativecommons.org/licenses/by/4.0/

Large language models (LLMs) have advanced to encompass extensive knowledge across diverse domains. Yet controlling what a large language model should not know is important for ensuring alignment and thus safe use. However, accurately and efficiently unlearning knowledge from an LLM remains challenging due to the potential collateral damage caused by the fuzzy boundary between retention and forgetting, and the large computational requirements for optimization across state-of-the-art models with hundreds of billions of parameters. In this work, we present \textbf{Embedding-COrrupted (ECO) Prompts}, a lightweight unlearning framework for large language models to address both the challenges of knowledge entanglement and unlearning efficiency. Instead of relying on the LLM itself to unlearn, we enforce an unlearned state during inference by employing a prompt classifier to identify and safeguard prompts to forget. We learn corruptions added to prompt embeddings via zeroth order optimization toward the unlearning objective offline and corrupt prompts flagged by the classifier during inference. We find that these embedding-corrupted prompts not only lead to desirable outputs that satisfy the unlearning objective but also closely approximate the output from a model that has never been trained on the data intended for forgetting. Through extensive experiments on unlearning, we demonstrate the superiority of our method in achieving promising unlearning at \textit{nearly zero side effects} in general domains and domains closely related to the unlearned ones. Additionally, we highlight the scalability of our method to 100 LLMs, ranging from 0.5B to 236B parameters, incurring no additional cost as the number of parameters increases. We have made our code publicly available at \url{https://github.com/chrisliu298/llm-unlearn-eco}.

2406.08589

Qing Wang

Qing Wang, Matthias Ihme, Cenk Gazen, Yi-Fan Chen, John Anderson

FireBench: A High-fidelity Ensemble Simulation Framework for Exploring Wildfire Behavior and Data-driven Modeling

physics.comp-ph

http://creativecommons.org/licenses/by/4.0/

Background. Wildfire research uses ensemble methods to analyze fire behaviors and assess uncertainties. Nonetheless, current research methods are either confined to simple models or complex simulations with limits. Modern computing tools could allow for efficient, high-fidelity ensemble simulations. Aims. This study proposes a high-fidelity ensemble wildfire simulation framework for studying wildfire behavior, ML tasks, fire-risk assessment, and uncertainty analysis. Methods. In this research, we present a simulation framework that integrates the Swirl-Fire large-eddy simulation tool for wildfire predictions with the Vizier optimization platform for automated run-time management of ensemble simulations and large-scale batch processing. All simulations are executed on tensor-processing units to enhance computational efficiency. Key results. A dataset of 117 simulations is created, each with 1.35 billion mesh points. The simulations are compared to existing experimental data and show good agreement in terms of fire rate of spread. Computations are done for fire acceleration, mean rate of spread, and fireline intensity. Conclusions. Strong coupling between these 2 parameters are observed for the fire spread and intermittency. A critical Froude number that delineates fires from plume-driven to convection-driven is identified and confirmed with literature observations. Implications. The ensemble simulation framework is efficient in facilitating parametric wildfire studies.

2406.08747

Cheng-Kuang Wu

Cheng-Kuang Wu, Zhi Rui Tam, Chieh-Yen Lin, Yun-Nung Chen, Hung-yi Lee

StreamBench: Towards Benchmarking Continuous Improvement of Language Agents

NeurIPS 2024 Track on Datasets and Benchmarks

cs.CL

http://creativecommons.org/licenses/by-sa/4.0/

Recent works have shown that large language model (LLM) agents are able to improve themselves from experience, which is an important ability for continuous enhancement post-deployment. However, existing benchmarks primarily evaluate their innate capabilities and do not assess their ability to improve over time. To address this gap, we introduce StreamBench, a pioneering benchmark designed to evaluate the continuous improvement of LLM agents over an input-feedback sequence. StreamBench simulates an online learning environment where LLMs receive a continuous flow of feedback stream and iteratively enhance their performance. In addition, we propose several simple yet effective baselines for improving LLMs on StreamBench, and provide a comprehensive analysis to identify critical components that contribute to successful streaming strategies. Our work serves as a stepping stone towards developing effective online learning strategies for LLMs, paving the way for more adaptive AI systems in streaming scenarios. Source code: https://github.com/stream-bench/stream-bench. Benchmark website: https://stream-bench.github.io.

2406.08845

Tianle Zhang

Tianle Zhang, Langtian Ma, Yuchen Yan, Yuchen Zhang, Kai Wang, Yue Yang, Ziyao Guo, Wenqi Shao, Yang You, Yu Qiao, Ping Luo, Kaipeng Zhang

Rethinking Human Evaluation Protocol for Text-to-Video Models: Enhancing Reliability,Reproducibility, and Practicality

cs.CV

http://creativecommons.org/licenses/by/4.0/

Recent text-to-video (T2V) technology advancements, as demonstrated by models such as Gen2, Pika, and Sora, have significantly broadened its applicability and popularity. Despite these strides, evaluating these models poses substantial challenges. Primarily, due to the limitations inherent in automatic metrics, manual evaluation is often considered a superior method for assessing T2V generation. However, existing manual evaluation protocols face reproducibility, reliability, and practicality issues. To address these challenges, this paper introduces the Text-to-Video Human Evaluation (T2VHE) protocol, a comprehensive and standardized protocol for T2V models. The T2VHE protocol includes well-defined metrics, thorough annotator training, and an effective dynamic evaluation module. Experimental results demonstrate that this protocol not only ensures high-quality annotations but can also reduce evaluation costs by nearly 50\%. We will open-source the entire setup of the T2VHE protocol, including the complete protocol workflow, the dynamic evaluation component details, and the annotation interface code. This will help communities establish more sophisticated human assessment protocols.

2406.08944

Yuki Tokushige

Yuki Tokushige

Graphical proof of Ginibre's inequality

7 pages, 1 figure, comments are welcome. deleted the last section

math.PR math-ph math.MP

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

In this short note, we will give a new combinatorial proof of Ginibre's inequality for XY models. Our proof is based on multigraph representations introduced by van Engelenburg-Lis (2023) and a new combinatorial bijection.

2406.09136

Chao Du

Xuan Zhang, Chao Du, Tianyu Pang, Qian Liu, Wei Gao, Min Lin

Chain of Preference Optimization: Improving Chain-of-Thought Reasoning in LLMs

NeurIPS 2024

cs.CL cs.LG

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

The recent development of chain-of-thought (CoT) decoding has enabled large language models (LLMs) to generate explicit logical reasoning paths for complex problem-solving. However, research indicates that these paths are not always deliberate and optimal. The tree-of-thought (ToT) method employs tree-searching to extensively explore the reasoning space and find better reasoning paths that CoT decoding might overlook. This deliberation, however, comes at the cost of significantly increased inference complexity. In this work, we demonstrate that fine-tuning LLMs leveraging the search tree constructed by ToT allows CoT to achieve similar or better performance, thereby avoiding the substantial inference burden. This is achieved through Chain of Preference Optimization (CPO), where LLMs are fine-tuned to align each step of the CoT reasoning paths with those of ToT using the inherent preference information in the tree-search process. Extensive experimental results show that CPO significantly improves LLM performance in solving a variety of complex problems, including question answering, fact verification, and arithmetic reasoning, demonstrating its effectiveness. Our code is available at https://github.com/sail-sg/CPO.

2406.09168

Soufiane Belharbi

Soufiane Belharbi, Mara KM Whitford, Phuong Hoang, Shakeeb Murtaza, Luke McCaffrey, Eric Granger

SR-CACO-2: A Dataset for Confocal Fluorescence Microscopy Image Super-Resolution

27 pages, 15 figures, NeurIPS 2024

eess.IV cs.CV cs.LG

http://creativecommons.org/licenses/by/4.0/

Confocal fluorescence microscopy is one of the most accessible and widely used imaging techniques for the study of biological processes at the cellular and subcellular levels. Scanning confocal microscopy allows the capture of high-quality images from thick three-dimensional (3D) samples, yet suffers from well-known limitations such as photobleaching and phototoxicity of specimens caused by intense light exposure, limiting its applications. Cellular damage can be alleviated by changing imaging parameters to reduce light exposure, often at the expense of image quality. Machine/deep learning methods for single-image super-resolution (SISR) can be applied to restore image quality by upscaling lower-resolution (LR) images to yield high-resolution images (HR). These SISR methods have been successfully applied to photo-realistic images due partly to the abundance of publicly available data. In contrast, the lack of publicly available data partly limits their application and success in scanning confocal microscopy. In this paper, we introduce a large scanning confocal microscopy dataset named SR-CACO-2 that is comprised of low- and high-resolution image pairs marked for three different fluorescent markers. It allows the evaluation of performance of SISR methods on three different upscaling levels (X2, X4, X8). SR-CACO-2 contains the human epithelial cell line Caco-2 (ATCC HTB-37), and it is composed of 2,200 unique images, captured with four resolutions and three markers, forming 9,937 image patches for SISR methods. We provide benchmarking results for 16 state-of-the-art methods of the main SISR families. Results show that these methods have limited success in producing high-resolution textures. The dataset is freely accessible under a Creative Commons license (CC BY-NC-SA 4.0). Our dataset, code and pretrained weights for SISR methods are available: https://github.com/sbelharbi/sr-caco-2.

2406.09451

Christopher Pulliam

Aaron J. Hadley and Christopher L. Pulliam

Enhancing Activity Recognition After Stroke: Generative Adversarial Networks for Kinematic Data Augmentation

13 pages, 5 figures, 3 tables; resubmitted to Sensors

Sensors 2024, 24(21), 6861

10.3390/s24216861

cs.LG eess.SP

http://creativecommons.org/licenses/by/4.0/

The generalizability of machine learning (ML) models for wearable monitoring in stroke rehabilitation is often constrained by the limited scale and heterogeneity of available data. Data augmentation addresses this challenge by adding computationally derived data to real data to enrich the variability represented in the training set. Traditional augmentation methods, such as rotation, permutation, and time-warping, have shown some benefits in improving classifier performance, but often fail to produce realistic training examples. This study employs Conditional Generative Adversarial Networks (cGANs) to create synthetic kinematic data from a publicly available dataset, closely mimicking the experimentally measured reaching movements of stroke survivors. This approach not only captures the complex temporal dynamics and common movement patterns after stroke, but also significantly enhances the training dataset. By training deep learning models on both synthetic and experimental data, we enhanced task classification accuracy: models incorporating synthetic data attained an overall accuracy of 80.0%, significantly higher than the 66.1% seen in models trained solely with real data. These improvements allow for more precise task classification, offering clinicians the potential to monitor patient progress more accurately and tailor rehabilitation interventions more effectively.

2406.09636

Jianfeng Chen

Jianfeng Chen, Sha Liu, Rui Zhang, Hao Jin, Congshan Zhuo, Ming Fang, Yanguang Yang, Chengwen Zhong

A gas-surface interaction algorithm for discrete velocity methods in predicting rarefied and multi-scale flows: For Cercignani-Lampis boundary model

physics.flu-dyn

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

The discrete velocity method (DVM) for rarefied flows and unified methods based on the DVM framework for flows in all regimes have worked well as precise flow solvers over the past decades and have been successfully extended to other important physical fields. However, these methods primarily focus on modeling gas-gas interactions. For gas-surface interactions (GSI) at the wall boundary, they usually use the full accommodation diffuse reflection model, which cannot accurately describe the behavior of reflected gas molecules in rarefied flows. To overcome this bottleneck and extend the DVM and unified methods to more realistic boundary conditions, a Cercignani-Lampis (CL) boundary with different momentum and thermal energy accommodations is proposed and integrated into the DVM framework. In this work, by giving the macroscopic flux from the numerical quadrature of the incident molecular distribution, the reflected macroscopic flux can be obtained for the given accommodation coefficients. Then, an anisotropic Gaussian distribution can be found for the reflected molecules, whose parameters are determined by the calculated reflected macroscopic flux. These macroscopic flux and microscopic Gaussian distribution form a complete physical process for the reflected molecules. Furthermore, the CL boundary is integrated into the unified gas-kinetic scheme (UGKS), making it suitable for the simulation of both monatomic and diatomic gas flows, and it accommodates both the conventional Cartesian velocity space and the recently developed efficient unstructured velocity space. Moreover, this new GSI boundary is suitable for both explicit and implicit schemes, offering better performance for flow prediction. Finally, the performance of the new boundary is validated through a series of numerical tests covering a wide range of Knudsen and Mach numbers.

2406.09705

Koji Ishiwata

Koji Ishiwata, Kentaro Nomura

Collective excitations in magnetic topological insulators and axion dark matter search

33 pages, 14 figures, published in JHEP

JHEP10(2024)225

10.1007/JHEP10(2024)225

KANAZAWA-24-06

hep-ph astro-ph.CO cond-mat.str-el

http://creativecommons.org/licenses/by/4.0/

We investigate collective excitations in magnetic topological insulators (TIs) and their impact on particle axion detection. In the three-dimensional TI model with the Hubbard term, the effective action of magnons and amplitude modes is formulated by dynamical susceptibility under the antiferromagnetic and ferromagnetic states. One of the amplitude modes is identified as ``axionic'' quasi-particle and its effective coupling to the electromagnetic fields turns out to be enhanced by about four orders of magnitude larger than the previous estimate, which may drastically change the sensitivity of the particle axion search using ``axion'' in magnetic TIs.

2406.09714

John Cherian

John J. Cherian, Isaac Gibbs, Emmanuel J. Cand\`es

Large language model validity via enhanced conformal prediction methods

24 pages, 14 figures, NeurIPS

stat.ML cs.LG stat.ME

http://creativecommons.org/licenses/by/4.0/

We develop new conformal inference methods for obtaining validity guarantees on the output of large language models (LLMs). Prior work in conformal language modeling identifies a subset of the text that satisfies a high-probability guarantee of correctness. These methods work by filtering claims from the LLM's original response if a scoring function evaluated on the claim fails to exceed a threshold calibrated via split conformal prediction. Existing methods in this area suffer from two deficiencies. First, the guarantee stated is not conditionally valid. The trustworthiness of the filtering step may vary based on the topic of the response. Second, because the scoring function is imperfect, the filtering step can remove many valuable and accurate claims. We address both of these challenges via two new conformal methods. First, we generalize the conditional conformal procedure of Gibbs et al. (2023) in order to adaptively issue weaker guarantees when they are required to preserve the utility of the output. Second, we show how to systematically improve the quality of the scoring function via a novel algorithm for differentiating through the conditional conformal procedure. We demonstrate the efficacy of our approach on biography and medical question-answering datasets.

2406.09867

Xingming Long

Xingming Long, Jie Zhang, Shiguang Shan, Xilin Chen

Rethinking the Evaluation of Out-of-Distribution Detection: A Sorites Paradox

Accepted at NeurIPS 2024

cs.CV

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Most existing out-of-distribution (OOD) detection benchmarks classify samples with novel labels as the OOD data. However, some marginal OOD samples actually have close semantic contents to the in-distribution (ID) sample, which makes determining the OOD sample a Sorites Paradox. In this paper, we construct a benchmark named Incremental Shift OOD (IS-OOD) to address the issue, in which we divide the test samples into subsets with different semantic and covariate shift degrees relative to the ID dataset. The data division is achieved through a shift measuring method based on our proposed Language Aligned Image feature Decomposition (LAID). Moreover, we construct a Synthetic Incremental Shift (Syn-IS) dataset that contains high-quality generated images with more diverse covariate contents to complement the IS-OOD benchmark. We evaluate current OOD detection methods on our benchmark and find several important insights: (1) The performance of most OOD detection methods significantly improves as the semantic shift increases; (2) Some methods like GradNorm may have different OOD detection mechanisms as they rely less on semantic shifts to make decisions; (3) Excessive covariate shifts in the image are also likely to be considered as OOD for some methods. Our code and data are released in https://github.com/qqwsad5/IS-OOD.

2406.09895

Argyro Damoulaki

Argyro Damoulaki, Ioannis Ntzoufras and Konstantinos Pelechrinis

Lasso Multinomial Performance Indicators for in-play Basketball Data

stat.AP

http://creativecommons.org/licenses/by/4.0/

A typical approach to quantify the contribution of each player in basketball uses the plus-minus method. The ratings obtained by such a method are estimated using simple regression models and their regularized variants, with response variable being either the points scored or the point differences. To capture more precisely the effect of each player, detailed possession-based play-by-play data may be used. This is the direction we take in this article, in which we investigate the performance of regularized adjusted plus-minus (RAPM) indicators estimated by different regularized models having as a response the number of points scored in each possession. Therefore, we use possession play-by-play data from all NBA games for the season 2021-22 (322,852 possessions). We initially present simple regression model-based indices starting from the implementation of ridge regression which is the standard technique in the relevant literature. We proceed with the lasso approach which has specific advantages and better performance than ridge regression when compared with selected objective validation criteria. Then, we implement regularized binary and multinomial logistic regression models to obtain more accurate performance indicators since the response is a discrete variable taking values mainly from zero to three. Our final proposal is an improved RAPM measure which is based on the expected points of a multinomial logistic regression model where each player's contribution is weighted by his participation in the team's possessions. The proposed indicator, called weighted expected points (wEPTS), outperforms all other RAPM measures we investigate in this study.

2406.10490

Ziyu Xu

Ziyu Xu, Nikos Karampatziakis, Paul Mineiro

Active, anytime-valid risk controlling prediction sets

22 pages, 3 figures. Accepted to NeurIPS 2024

stat.ML cs.LG

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Rigorously establishing the safety of black-box machine learning models concerning critical risk measures is important for providing guarantees about model behavior. Recently, Bates et. al. (JACM '24) introduced the notion of a risk controlling prediction set (RCPS) for producing prediction sets that are statistically guaranteed low risk from machine learning models. Our method extends this notion to the sequential setting, where we provide guarantees even when the data is collected adaptively, and ensures that the risk guarantee is anytime-valid, i.e., simultaneously holds at all time steps. Further, we propose a framework for constructing RCPSes for active labeling, i.e., allowing one to use a labeling policy that chooses whether to query the true label for each received data point and ensures that the expected proportion of data points whose labels are queried are below a predetermined label budget. We also describe how to use predictors (i.e., the machine learning model for which we provide risk control guarantees) to further improve the utility of our RCPSes by estimating the expected risk conditioned on the covariates. We characterize the optimal choices of label policy and predictor under a fixed label budget and show a regret result that relates the estimation error of the optimal labeling policy and predictor to the wealth process that underlies our RCPSes. Lastly, we present practical ways of formulating label policies and empirically show that our label policies use fewer labels to reach higher utility than naive baseline labeling strategies on both simulations and real data.

2406.10796

Gabriel Guo

Gabe Guo, Tristan Saidi, Maxwell Terban, Michele Valsecchi, Simon JL Billinge, Hod Lipson

Ab Initio Structure Solutions from Nanocrystalline Powder Diffraction Data

physics.comp-ph cond-mat.mes-hall cs.AI cs.LG

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

A major challenge in materials science is the determination of the structure of nanometer sized objects. Here we present a novel approach that uses a generative machine learning model based on diffusion processes that is trained on 45,229 known structures. The model factors both the measured diffraction pattern as well as relevant statistical priors on the unit cell of atomic cluster structures. Conditioned only on the chemical formula and the information-scarce finite-size broadened powder diffraction pattern, we find that our model, PXRDnet, can successfully solve simulated nanocrystals as small as 10 angstroms across 200 materials of varying symmetry and complexity, including structures from all seven crystal systems. We show that our model can successfully and verifiably determine structural candidates four out of five times, with average error among these candidates being only 7% (as measured by post-Rietveld refinement R-factor). Furthermore, PXRDnet is capable of solving structures from noisy diffraction patterns gathered in real-world experiments. We suggest that data driven approaches, bootstrapped from theoretical simulation, will ultimately provide a path towards determining the structure of previously unsolved nano-materials.

2406.11174

Fatih Efe Bilgen

Fatih E. Bilgen and Ozgur B. Akan

Mycorrhizal Fungi and Plant Symbiosis for Energy Harvesting in the Internet of Plants

10 pages, 5 figures, conference

NANOCOM '24: The 11th Annual ACM International Conference on Nanoscale Computing and Communication Milan Italy October 28 - 30, 2024

10.1145/3686015.3689365

cs.ET eess.SP

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Biological entities in nature have developed sophisticated communication methods over millennia to facilitate cooperation. Among these entities, plants are some of the most intricate communicators. They interact with each other through various communication modalities, creating networks that enable the exchange of information and nutrients. In this paper, we explore this collective behavior and its components. We then introduce the concept of agent plants, outlining their architecture and detailing the tasks of each unit. Additionally, we investigate the mycorrhizal fungi-plant symbiosis to extract glucose for energy harvesting. We propose an architecture that converts the chemical energy stored in these glucose molecules into electrical energy. We conduct comprehensive analyses of the proposed architecture to validate its effectiveness.

2406.11182

Jounghun Lee

Jounghun Lee, Jun-Sung Moon (Seoul National University)

The dependence of halo bias on the protohalo shape alignment with the initial tidal field

Accepted for publication in JCAP, 7 figures, a typo in eq.(1) fixed, minor revision after referee's comments

JCAP 10 (2024) 102

astro-ph.CO astro-ph.GA

http://creativecommons.org/licenses/by/4.0/

We present a numerical evidence supporting the primordial origin of secondary halo bias even on the galactic mass scale. Analyzing the data from the IllustrisTNG 300-1 simulations, we investigate the dependence of halo bias on the degree of misalignment between the protohalo inertia and initial tidal tensors, $\tau$, measured at redshift, $z_{i}=127$. From the TNG 300-1 galactic halos in logarithmic mass range of $10.5< m\equiv \log[M/(h^{-1}M_{\odot})]\le 13$ identified at $z=0,\ 0.5$ and $1$, a clear signal of $\tau$ bias is detected. For the case that $\tau$ is measured from the initial tidal field smoothed on the scale of $R_{f}/(h^{-1}\,{\rm Mpc})\lesssim 1$, the halo $\tau$ bias is found to be very similar in its tendency and amplitude to the spin bias at all of the three redshifts, if the effects of backsplash halos are properly eliminated. For the case of $R_{f}/(h^{-1}\,{\rm Mpc})=2$, the $\tau$ bias at $z=1$ turns out to behave like the age bias, diminishing rapidly in the range of $m> 12$. At $z=0$ and $0.5$, however, the $\tau$ and age bias factors show large differences in their overall strengths, which is attributed to the dominant nonlinear effects that undermine the former but enhance the latter. Given these numerical results along with the previous finding that $\tau$ shares a large amount of mutual information with the formation epochs and spin parameters of galactic halos, it is concluded that the origins of halo age and spin bias must be closely linked with the primordial factor, $\tau$, and that the difference in the tendency between the two bias factors on the galactic mass scale reflects the multi-scale influence of $\tau$ on the halo secondary properties.

2406.11200

Shirley Wu

Shirley Wu, Shiyu Zhao, Qian Huang, Kexin Huang, Michihiro Yasunaga, Kaidi Cao, Vassilis N. Ioannidis, Karthik Subbian, Jure Leskovec, James Zou

AvaTaR: Optimizing LLM Agents for Tool Usage via Contrastive Reasoning

NeurIPS 2024 main conference

cs.LG cs.CL

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Large language model (LLM) agents have demonstrated impressive capabilities in utilizing external tools and knowledge to boost accuracy and reduce hallucinations. However, developing prompting techniques that enable LLM agents to effectively use these tools and knowledge remains a heuristic and labor-intensive task. Here, we introduce AvaTaR, a novel and automated framework that optimizes an LLM agent to effectively leverage provided tools, improving performance on a given task. During optimization, we design a comparator module to iteratively deliver insightful and comprehensive prompts to the LLM agent by contrastively reasoning between positive and negative examples sampled from training data. We demonstrate AvaTaR on four complex multimodal retrieval datasets featuring textual, visual, and relational information, and three general question-answering (QA) datasets. We find AvaTaR consistently outperforms state-of-the-art approaches across all seven tasks, exhibiting strong generalization ability when applied to novel cases and achieving an average relative improvement of 14% on the Hit@1 metric for the retrieval datasets and 13% for the QA datasets. Code and dataset are available at https://github.com/zou-group/avatar.

2406.11250

Yuxia Wang

Muhammad Arslan Manzoor, Yuxia Wang, Minghan Wang, Preslav Nakov

Can Machines Resonate with Humans? Evaluating the Emotional and Empathic Comprehension of LMs

19 pages

cs.CL

http://creativecommons.org/licenses/by/4.0/

Empathy plays a pivotal role in fostering prosocial behavior, often triggered by the sharing of personal experiences through narratives. However, modeling empathy using NLP approaches remains challenging due to its deep interconnection with human interaction dynamics. Previous approaches, which involve fine-tuning language models (LMs) on human-annotated empathic datasets, have had limited success. In our pursuit of improving empathy understanding in LMs, we propose several strategies, including contrastive learning with masked LMs and supervised fine-tuning with large language models. While these methods show improvements over previous methods, the overall results remain unsatisfactory. To better understand this trend, we performed an analysis which reveals a low agreement among annotators. This lack of consensus hinders training and highlights the subjective nature of the task. We also explore the cultural impact on annotations. To study this, we meticulously collected story pairs in Urdu language and find that subjectivity in interpreting empathy among annotators appears to be independent of cultural background. Our systematic exploration of LMs' understanding of empathy reveals substantial opportunities for further investigation in both task formulation and modeling.

2406.11271

Anas Awadalla

Anas Awadalla, Le Xue, Oscar Lo, Manli Shu, Hannah Lee, Etash Kumar Guha, Matt Jordan, Sheng Shen, Mohamed Awadalla, Silvio Savarese, Caiming Xiong, Ran Xu, Yejin Choi, Ludwig Schmidt

MINT-1T: Scaling Open-Source Multimodal Data by 10x: A Multimodal Dataset with One Trillion Tokens

cs.CV cs.LG

http://creativecommons.org/licenses/by/4.0/

Multimodal interleaved datasets featuring free-form interleaved sequences of images and text are crucial for training frontier large multimodal models (LMMs). Despite the rapid progression of open-source LMMs, there remains a pronounced scarcity of large-scale, diverse open-source multimodal interleaved datasets. In response, we introduce MINT-1T, the most extensive and diverse open-source Multimodal INTerleaved dataset to date. MINT-1T comprises one trillion text tokens and 3.4 billion images, a 10x scale-up from existing open-source datasets. Additionally, we include previously untapped sources such as PDFs and ArXiv papers. As scaling multimodal interleaved datasets requires substantial engineering effort, sharing the data curation process and releasing the dataset greatly benefits the community. Our experiments show that LMMs trained on MINT-1T rival the performance of models trained on the previous leading dataset, OBELICS. Our data and code will be released at https://github.com/mlfoundations/MINT-1T.

2406.11717

Andy Arditi

Andy Arditi, Oscar Obeso, Aaquib Syed, Daniel Paleka, Nina Panickssery, Wes Gurnee, Neel Nanda

Refusal in Language Models Is Mediated by a Single Direction

cs.LG cs.AI cs.CL

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Conversational large language models are fine-tuned for both instruction-following and safety, resulting in models that obey benign requests but refuse harmful ones. While this refusal behavior is widespread across chat models, its underlying mechanisms remain poorly understood. In this work, we show that refusal is mediated by a one-dimensional subspace, across 13 popular open-source chat models up to 72B parameters in size. Specifically, for each model, we find a single direction such that erasing this direction from the model's residual stream activations prevents it from refusing harmful instructions, while adding this direction elicits refusal on even harmless instructions. Leveraging this insight, we propose a novel white-box jailbreak method that surgically disables refusal with minimal effect on other capabilities. Finally, we mechanistically analyze how adversarial suffixes suppress propagation of the refusal-mediating direction. Our findings underscore the brittleness of current safety fine-tuning methods. More broadly, our work showcases how an understanding of model internals can be leveraged to develop practical methods for controlling model behavior.

2406.11779

Jason Gross

Jason Gross, Rajashree Agrawal, Thomas Kwa, Euan Ong, Chun Hei Yip, Alex Gibson, Soufiane Noubir, Lawrence Chan

Compact Proofs of Model Performance via Mechanistic Interpretability

accepted to the 38th Conference on Neural Information Processing Systems (NeurIPS 2024)

cs.LG cs.LO

http://creativecommons.org/licenses/by/4.0/

We propose using mechanistic interpretability -- techniques for reverse engineering model weights into human-interpretable algorithms -- to derive and compactly prove formal guarantees on model performance. We prototype this approach by formally proving accuracy lower bounds for a small transformer trained on Max-of-K, validating proof transferability across 151 random seeds and four values of K. We create 102 different computer-assisted proof strategies and assess their length and tightness of bound on each of our models. Using quantitative metrics, we find that shorter proofs seem to require and provide more mechanistic understanding. Moreover, we find that more faithful mechanistic understanding leads to tighter performance bounds. We confirm these connections by qualitatively examining a subset of our proofs. Finally, we identify compounding structureless errors as a key challenge for using mechanistic interpretability to generate compact proofs on model performance.

2406.11906

Jingbo Zhou

Jingbo Zhou, Shaorong Chen, Jun Xia, Sizhe Liu, Tianze Ling, Wenjie Du, Yue Liu, Jianwei Yin, Stan Z. Li

NovoBench: Benchmarking Deep Learning-based De Novo Peptide Sequencing Methods in Proteomics

NeurIPS 2024 D&B track

q-bio.QM cs.AI

http://creativecommons.org/licenses/by/4.0/

Tandem mass spectrometry has played a pivotal role in advancing proteomics, enabling the high-throughput analysis of protein composition in biological tissues. Many deep learning methods have been developed for \emph{de novo} peptide sequencing task, i.e., predicting the peptide sequence for the observed mass spectrum. However, two key challenges seriously hinder the further advancement of this important task. Firstly, since there is no consensus for the evaluation datasets, the empirical results in different research papers are often not comparable, leading to unfair comparison. Secondly, the current methods are usually limited to amino acid-level or peptide-level precision and recall metrics. In this work, we present the first unified benchmark NovoBench for \emph{de novo} peptide sequencing, which comprises diverse mass spectrum data, integrated models, and comprehensive evaluation metrics. Recent impressive methods, including DeepNovo, PointNovo, Casanovo, InstaNovo, AdaNovo and $\pi$-HelixNovo are integrated into our framework. In addition to amino acid-level and peptide-level precision and recall, we evaluate the models' performance in terms of identifying post-tranlational modifications (PTMs), efficiency and robustness to peptide length, noise peaks and missing fragment ratio, which are important influencing factors while seldom be considered. Leveraging this benchmark, we conduct a large-scale study of current methods, report many insightful findings that open up new possibilities for future development.

2406.11965

Michael M\"uller

Michael M\"uller, Suvodip Mukherjee, Geoffrey Ryan

Be careful in multi-messenger inference of the Hubble constant: A path forward for robust inference

21 pages, 10 figures, accepted in ApJL on October 24, 2024

astro-ph.CO astro-ph.HE gr-qc

http://creativecommons.org/licenses/by/4.0/

Multi-messenger observations of coalescing binary neutron stars (BNSs) are a direct probe of the expansion history of the universe and carry the potential to shed light on the disparity between low- and high-redshift measurements of the Hubble constant $H_0$. To measure the value of $H_0$ with such observations requires pristine inference of the luminosity distance and the true source redshift with minimal impact from systematics. In this analysis, we carry out joint inference on mock gravitational wave (GW) signals and their electromagnetic (EM) afterglows from BNS coalescences and find that the inclination angle inferred from the afterglow light curve and apparent superluminal motion can be precise, but need not be accurate and is subject to systematic uncertainty that could be as large as $1.5\sigma$. This produces a disparity between the EM and GW inferred inclination angles, which if not carefully treated when combining observations can bias the inferred value of $H_0$. We also find that already small misalignments of $3^{\circ}-6^{\circ}$ between the inherent system inclinations for the GW and EM emission can bias the inference by $\mathcal{O}(1-2\sigma)$ if not taken into account. As multi-messenger BNS observations are rare, we must make the most out of a small number of events and harness the increased precision, while avoiding reduced accuracy. We demonstrate how to mitigate these potential sources of bias by jointly inferring the mismatch between the GW- and EM-based inclination angles and $H_0$.

2406.12095

Letian Wang

Letian Wang, Seung Wook Kim, Jiawei Yang, Cunjun Yu, Boris Ivanovic, Steven L. Waslander, Yue Wang, Sanja Fidler, Marco Pavone, Peter Karkus

DistillNeRF: Perceiving 3D Scenes from Single-Glance Images by Distilling Neural Fields and Foundation Model Features

Accepted by Advances in Neural Information Processing Systems (NeurIPS 2024)

cs.CV cs.AI cs.RO

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

We propose DistillNeRF, a self-supervised learning framework addressing the challenge of understanding 3D environments from limited 2D observations in outdoor autonomous driving scenes. Our method is a generalizable feedforward model that predicts a rich neural scene representation from sparse, single-frame multi-view camera inputs with limited view overlap, and is trained self-supervised with differentiable rendering to reconstruct RGB, depth, or feature images. Our first insight is to exploit per-scene optimized Neural Radiance Fields (NeRFs) by generating dense depth and virtual camera targets from them, which helps our model to learn enhanced 3D geometry from sparse non-overlapping image inputs. Second, to learn a semantically rich 3D representation, we propose distilling features from pre-trained 2D foundation models, such as CLIP or DINOv2, thereby enabling various downstream tasks without the need for costly 3D human annotations. To leverage these two insights, we introduce a novel model architecture with a two-stage lift-splat-shoot encoder and a parameterized sparse hierarchical voxel representation. Experimental results on the NuScenes and Waymo NOTR datasets demonstrate that DistillNeRF significantly outperforms existing comparable state-of-the-art self-supervised methods for scene reconstruction, novel view synthesis, and depth estimation; and it allows for competitive zero-shot 3D semantic occupancy prediction, as well as open-world scene understanding through distilled foundation model features. Demos and code will be available at https://distillnerf.github.io/.

2406.12303

Yiheng Li

Yiheng Li, Heyang Jiang, Akio Kodaira, Masayoshi Tomizuka, Kurt Keutzer, Chenfeng Xu

Immiscible Diffusion: Accelerating Diffusion Training with Noise Assignment

cs.CV

http://creativecommons.org/licenses/by/4.0/

In this paper, we point out that suboptimal noise-data mapping leads to slow training of diffusion models. During diffusion training, current methods diffuse each image across the entire noise space, resulting in a mixture of all images at every point in the noise layer. We emphasize that this random mixture of noise-data mapping complicates the optimization of the denoising function in diffusion models. Drawing inspiration from the immiscibility phenomenon in physics, we propose Immiscible Diffusion, a simple and effective method to improve the random mixture of noise-data mapping. In physics, miscibility can vary according to various intermolecular forces. Thus, immiscibility means that the mixing of molecular sources is distinguishable. Inspired by this concept, we propose an assignment-then-diffusion training strategy to achieve Immiscible Diffusion. As one example, prior to diffusing the image data into noise, we assign diffusion target noise for the image data by minimizing the total image-noise pair distance in a mini-batch. The assignment functions analogously to external forces to expel the diffuse-able areas of images, thus mitigating the inherent difficulties in diffusion training. Our approach is remarkably simple, requiring only one line of code to restrict the diffuse-able area for each image while preserving the Gaussian distribution of noise. In this way, each image is preferably projected to nearby noise. Experiments demonstrate that our method can achieve up to 3x faster training for unconditional Consistency Models on the CIFAR dataset, as well as for DDIM and Stable Diffusion on CelebA and ImageNet dataset, and in class-conditional training and fine-tuning. In addition, we conducted a thorough analysis that sheds light on how it improves diffusion training speed while improving fidelity. The code is available at https://yhli123.github.io/immiscible-diffusion

2406.12747

Wenjie Du

Wenjie Du, Jun Wang, Linglong Qian, Yiyuan Yang, Zina Ibrahim, Fanxing Liu, Zepu Wang, Haoxin Liu, Zhiyuan Zhao, Yingjie Zhou, Wenjia Wang, Kaize Ding, Yuxuan Liang, B. Aditya Prakash, Qingsong Wen

TSI-Bench: Benchmarking Time Series Imputation

cs.LG cs.AI

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Effective imputation is a crucial preprocessing step for time series analysis. Despite the development of numerous deep learning algorithms for time series imputation, the community lacks standardized and comprehensive benchmark platforms to effectively evaluate imputation performance across different settings. Moreover, although many deep learning forecasting algorithms have demonstrated excellent performance, whether their modelling achievements can be transferred to time series imputation tasks remains unexplored. To bridge these gaps, we develop TSI-Bench, the first (to our knowledge) comprehensive benchmark suite for time series imputation utilizing deep learning techniques. The TSI-Bench pipeline standardizes experimental settings to enable fair evaluation of imputation algorithms and identification of meaningful insights into the influence of domain-appropriate missing rates and patterns on model performance. Furthermore, TSI-Bench innovatively provides a systematic paradigm to tailor time series forecasting algorithms for imputation purposes. Our extensive study across 34,804 experiments, 28 algorithms, and 8 datasets with diverse missingness scenarios demonstrates TSI-Bench's effectiveness in diverse downstream tasks and potential to unlock future directions in time series imputation research and analysis. All source code and experiment logs are released at https://github.com/WenjieDu/AwesomeImputation.

2406.13074

Guillaume Qu\'etant

Guillaume Qu\'etant, John Andrew Raine, Matthew Leigh, Debajyoti Sengupta, Tobias Golling

PIPPIN: Generating variable length full events from partons

Phys. Rev. D 110, 076023 (Published 21 October 2024)

10.1103/PhysRevD.110.076023

hep-ph cs.LG hep-ex stat.ML

http://creativecommons.org/licenses/by/4.0/

This paper presents a novel approach for directly generating full events at detector-level from parton-level information, leveraging cutting-edge machine learning techniques. To address the challenge of multiplicity variations between parton and reconstructed object spaces, we employ transformers, score-based models and normalizing flows. Our method tackles the inherent complexities of the stochastic transition between these two spaces and achieves remarkably accurate results. The combination of innovative techniques and the achieved accuracy demonstrates the potential of our approach in advancing the field and opens avenues for further exploration. This research contributes to the ongoing efforts in high-energy physics and generative modelling, providing a promising direction for enhanced precision in fast detector simulation.

2406.13185

Yingzhe Peng

Yingzhe Peng, Chenduo Hao, Xu Yang, Jiawei Peng, Xinting Hu, Xin Geng

LIVE: Learnable In-Context Vector for Visual Question Answering

cs.CL

http://creativecommons.org/licenses/by/4.0/

As language models continue to scale, Large Language Models (LLMs) have exhibited emerging capabilities in In-Context Learning (ICL), enabling them to solve language tasks by prefixing a few in-context demonstrations (ICDs) as context. Inspired by these advancements, researchers have extended these techniques to develop Large Multimodal Models (LMMs) with ICL capabilities. However, applying ICL usually faces two major challenges: 1) using more ICDs will largely increase the inference time and 2) the performance is sensitive to the selection of ICDs. These challenges are further exacerbated in LMMs due to the integration of multiple data types and the combinational complexity of multimodal ICDs. Recently, to address these challenges, some NLP studies introduce non-learnable In-Context Vectors (ICVs) which extract useful task information from ICDs into a single vector and then insert it into the LLM to help solve the corresponding task. However, although useful in simple NLP tasks, these non-learnable methods fail to handle complex multimodal tasks like Visual Question Answering (VQA). In this study, we propose Learnable In-Context VEctor (LIVE) to distill essential task information from demonstrations, improving ICL performance in LMMs. Experiments show that LIVE can significantly reduce computational costs while enhancing accuracy in VQA tasks compared to traditional ICL and other non-learnable ICV methods. The code is available at \url{https://github.com/ForJadeForest/LIVE-Learnable-In-Context-Vector}.

2406.13495

Zhiyuan Yan

Zhiyuan Yan, Taiping Yao, Shen Chen, Yandan Zhao, Xinghe Fu, Junwei Zhu, Donghao Luo, Chengjie Wang, Shouhong Ding, Yunsheng Wu, Li Yuan

DF40: Toward Next-Generation Deepfake Detection

arXiv admin note: text overlap with arXiv:2108.05080 by other authors

cs.CV

http://creativecommons.org/licenses/by-nc-nd/4.0/

We propose a new comprehensive benchmark to revolutionize the current deepfake detection field to the next generation. Predominantly, existing works identify top-notch detection algorithms and models by adhering to the common practice: training detectors on one specific dataset (e.g., FF++) and testing them on other prevalent deepfake datasets. This protocol is often regarded as a "golden compass" for navigating SoTA detectors. But can these stand-out "winners" be truly applied to tackle the myriad of realistic and diverse deepfakes lurking in the real world? If not, what underlying factors contribute to this gap? In this work, we found the dataset (both train and test) can be the "primary culprit" due to: (1) forgery diversity: Deepfake techniques are commonly referred to as both face forgery and entire image synthesis. Most existing datasets only contain partial types of them, with limited forgery methods implemented; (2) forgery realism: The dominated training dataset, FF++, contains out-of-date forgery techniques from the past four years. "Honing skills" on these forgeries makes it difficult to guarantee effective detection generalization toward nowadays' SoTA deepfakes; (3) evaluation protocol: Most detection works perform evaluations on one type, which hinders the development of universal deepfake detectors. To address this dilemma, we construct a highly diverse deepfake detection dataset called DF40, which comprises 40 distinct deepfake techniques. We then conduct comprehensive evaluations using 4 standard evaluation protocols and 8 representative detection methods, resulting in over 2,000 evaluations. Through these evaluations, we provide an extensive analysis from various perspectives, leading to 7 new insightful findings. We also open up 4 valuable yet previously underexplored research questions to inspire future works. Our project page is https://github.com/YZY-stack/DF40.

2406.13740

Joel I-Jan Wang

Miuko Tanaka, Joel \^I-j. Wang, Thao H. Dinh, Daniel Rodan-Legrain, Sameia Zaman, Max Hays, Bharath Kannan, Aziza Almanakly, David K. Kim, Bethany M. Niedzielski, Kyle Serniak, Mollie E. Schwartz, Kenji Watanabe, Takashi Taniguchi, Jeffrey A. Grover, Terry P. Orlando, Simon Gustavsson, Pablo Jarillo-Herrero, and William D. Oliver

Superfluid Stiffness and Flat-Band Superconductivity in Magic-Angle Graphene Probed by cQED

cond-mat.supr-con cond-mat.mes-hall cond-mat.str-el quant-ph

http://arxiv.org/licenses/nonexclusive-distrib/1.0/

The physics of superconductivity in magic-angle twisted bilayer graphene (MATBG) is a topic of keen interest in moir\'e systems research, and it may provide insight into the pairing mechanism of other strongly correlated materials such as high-$T_{\mathrm{c}}$ superconductors. Here, we use DC-transport and microwave circuit quantum electrodynamics (cQED) to measure directly the superfluid stiffness of superconducting MATBG via its kinetic inductance. We find the superfluid stiffness to be much larger than expected from conventional Fermi liquid theory; rather, it is comparable to theoretical predictions involving quantum geometric effects that are dominant at the magic angle. The temperature dependence of the superfluid stiffness follows a power-law, which contraindicates an isotropic BCS model; instead, the extracted power-law exponents indicate an anisotropic superconducting gap, whether interpreted within the Fermi liquid framework or by considering quantum geometry of flat-band superconductivity. Moreover, a quadratic dependence of the superfluid stiffness on both DC and microwave current is observed, which is consistent with Ginzburg-Landau theory. Taken together, our findings indicate that MATBG is an unconventional superconductor with an anisotropic gap and strongly suggest a connection between quantum geometry, superfluid stiffness, and unconventional superconductivity in MATBG. The combined DC-microwave measurement platform used here is applicable to the investigation of other atomically thin superconductors.

2406.13762

Tan Nguyen

Rachel S.Y. Teo, Tan M. Nguyen

Unveiling the Hidden Structure of Self-Attention via Kernel Principal Component Analysis

10 pages in the main text. Published at NeurIPS 2024. The code is available at https://github.com/rachtsy/KPCA_code

cs.LG cs.AI cs.CL cs.CV stat.ML

http://creativecommons.org/licenses/by/4.0/

The remarkable success of transformers in sequence modeling tasks, spanning various applications in natural language processing and computer vision, is attributed to the critical role of self-attention. Similar to the development of most deep learning models, the construction of these attention mechanisms relies on heuristics and experience. In our work, we derive self-attention from kernel principal component analysis (kernel PCA) and show that self-attention projects its query vectors onto the principal component axes of its key matrix in a feature space. We then formulate the exact formula for the value matrix in self-attention, theoretically and empirically demonstrating that this value matrix captures the eigenvectors of the Gram matrix of the key vectors in self-attention. Leveraging our kernel PCA framework, we propose Attention with Robust Principal Components (RPC-Attention), a novel class of robust attention that is resilient to data contamination. We empirically demonstrate the advantages of RPC-Attention over softmax attention on the ImageNet-1K object classification, WikiText-103 language modeling, and ADE20K image segmentation task.