title stringlengths 14 183 | paper_url stringlengths 42 42 | authors listlengths 1 36 | type stringclasses 3
values | primary_area stringclasses 13
values | abstract large_stringlengths 246 3.6k | keywords listlengths 1 28 | TL;DR large_stringlengths 2 250 ⌀ | submission_number int64 1 29.3k | arxiv_id stringlengths 10 10 ⌀ | arxiv_id_source stringclasses 2
values | embedding listlengths 768 768 |
|---|---|---|---|---|---|---|---|---|---|---|---|
Time-o1: Time-Series Forecasting Needs Transformed Label Alignment | https://openreview.net/forum?id=RxWILaXuhb | [
"Hao Wang",
"Licheng Pan",
"Zhichao Chen",
"Xu Chen",
"Qingyang Dai",
"Lei Wang",
"Haoxuan Li",
"Zhouchen Lin"
] | Poster | applications | Training time-series forecasting models poses unique challenges in loss function design. Most existing approaches adopt temporal mean squared error, but this study reveals two critical limitations: (1) it ignores the presence of label autocorrelation, which biases it from the true label sequence likelihood; (2) it inv... | [
"Time-Series",
"Label Autocorrelation",
"Orthogonalization"
] | Learning to forecast in the transformed domain improves forecasting performance. | 29,297 | 2505.17847 | title_snapshot | [
-0.010552695952355862,
-0.039074987173080444,
-0.013647916726768017,
0.00856844149529934,
0.05646183341741562,
0.0523507297039032,
0.009915128350257874,
0.029793374240398407,
-0.006123242434114218,
-0.05104382336139679,
0.0006754604401066899,
0.020093517377972603,
-0.0781569704413414,
0.00... |
REVE: A Foundation Model for EEG - Adapting to Any Setup with Large-Scale Pretraining on 25,000 Subjects | https://openreview.net/forum?id=ZeFMtRBy4Z | [
"Yassine El Ouahidi",
"Jonathan Lys",
"Philipp Thölke",
"Nicolas Farrugia",
"Bastien Pasdeloup",
"Vincent Gripon",
"Karim Jerbi",
"Giulia Lioi"
] | Poster | neuroscience_and_cognitive_science | Foundation models have transformed AI by reducing reliance on task-specific data through large-scale pretraining. While successful in language and vision, their adoption in EEG has lagged due to the heterogeneity of public datasets, which are collected under varying protocols, devices, and electrode configurations. Exi... | [
"Foundation Model",
"EEG",
"SSL",
"BCI"
] | A scalable EEG foundation model leveraging 60,000+ hours of data, adaptable to any electrode setup, offering ready-to-use embeddings and state-of-the-art performance across diverse tasks. | 29,260 | 2510.21585 | title_snapshot | [
-0.001960654743015766,
-0.01708337292075157,
0.03870749473571777,
-0.009951246902346611,
0.03557745739817619,
0.017963062971830368,
0.04160674661397934,
0.027642477303743362,
-0.02467745542526245,
-0.06200456991791725,
0.0040960777550935745,
0.022838391363620758,
-0.059277284890413284,
-0.... |
ModHiFi: Identifying High Fidelity predictive components for Model Modification | https://openreview.net/forum?id=lClK4uBxSG | [
"Dhruva Kashyap",
"Chaitanya Murti",
"Pranav K Nayak",
"Tanay Narshana",
"Chiranjib Bhattacharyya"
] | Spotlight | deep_learning | Open weight models, which are ubiquitous, rarely provide access to their training data or loss function. This makes modifying such models for tasks such as pruning or unlearning, which are constrained by this unavailability, an active area of research. Existing techniques typically require gradients or ground-truth lab... | [
"Pruning",
"Machine Unlearning"
] | null | 29,227 | 2511.19566 | title_snapshot | [
-0.007194813806563616,
-0.050555240362882614,
0.007771169766783714,
0.04874604567885399,
0.05473264679312706,
0.032605282962322235,
0.01256540883332491,
0.011350243352353573,
-0.033389024436473846,
-0.03521822765469551,
-0.022047191858291626,
0.046933889389038086,
-0.06720126420259476,
0.0... |
The Structure of Relation Decoding Linear Operators in Large Language Models | https://openreview.net/forum?id=XsBzmJzJ2l | [
"Miranda Anna Christ",
"Adrián Csiszárik",
"Gergely Becsó",
"Dániel Varga"
] | Spotlight | deep_learning | This paper investigates the structure of linear operators introduced in Hernandez et al. [2023] that decode specific relational facts in transformer language models. We extend their single-relation findings to a collection of relations and systematically chart their organization. We show that such collections of relati... | [
"large language models",
"relations",
"tensor networks",
"interpretability"
] | We investigate the structure of relations in large language models, and compress linear relation decoding operators with tensor networks | 29,206 | 2510.26543 | title_snapshot | [
-0.02450931817293167,
-0.003615762572735548,
0.010746886022388935,
0.037720005959272385,
0.03907686844468117,
0.0394926443696022,
0.023779232054948807,
0.009504367597401142,
0.006521659437566996,
-0.0011351066641509533,
-0.029886143282055855,
0.036358095705509186,
-0.07387229800224304,
0.0... |
Vulnerable Data-Aware Adversarial Training | https://openreview.net/forum?id=yrrU5YChQr | [
"Yuqi Feng",
"Jiahao Fan",
"Yanan Sun"
] | Poster | deep_learning | Fast adversarial training (FAT) has been considered as one of the most effective alternatives to the computationally-intensive adversarial training. Generally, FAT methods pay equal attention to each sample of the target task. However, the distance between each sample and the decision boundary is different, learning sa... | [
"Adversarial Training",
"Adversarial Robustness",
"Decision Boundary Analysis"
] | null | 29,190 | null | null | [
0.0025452771224081516,
-0.04288822412490845,
0.0188825111836195,
0.07822257280349731,
0.03281926363706589,
0.020946605131030083,
0.03589266911149025,
-0.037991128861904144,
-0.013504552654922009,
-0.05184222012758255,
-0.014382720924913883,
-0.004456531722098589,
-0.06704135239124298,
-0.0... |
Tight analyses of first-order methods with error feedback | https://openreview.net/forum?id=hlPk6Hi43e | [
"Daniel Berg Thomsen",
"Adrien Taylor",
"Aymeric Dieuleveut"
] | Poster | optimization | Communication between agents often constitutes a major computational bottleneck in distributed learning. One of the most common mitigation strategies is to compress the information exchanged, thereby reducing communication overhead. To counteract the degradation in convergence associated with compressed communication, ... | [
"distributed optimization",
"distributed learning",
"error feedback",
"EF",
"EF21",
"tight analysis",
"performance estimation",
"convex optimization",
"large-scale machine learning"
] | null | 29,188 | 2506.05271 | title_snapshot | [
-0.007419840432703495,
-0.021904736757278442,
0.0030202006455510855,
0.048496413975954056,
0.044785015285015106,
0.04721665009856224,
0.02010035701096058,
-0.01673376001417637,
-0.020610207691788673,
-0.05540520325303078,
0.017499560490250587,
0.0064974636770784855,
-0.08215444535017014,
-... |
Cost-Sensitive Freeze-thaw Bayesian Optimization for Efficient Hyperparameter Tuning | https://openreview.net/forum?id=ZUb4JpNoJe | [
"Dong Bok Lee",
"Aoxuan Silvia Zhang",
"Byungjoo Kim",
"Junhyeon Park",
"Steven Adriaensen",
"Juho Lee",
"Sung Ju Hwang",
"Hae Beom Lee"
] | Poster | deep_learning | In this paper, we address the problem of cost-sensitive hyperparameter optimization (HPO) built upon freeze-thaw Bayesian optimization (BO). Specifically, we assume a scenario where users want to early-stop the HPO process when the expected performance improvement is not satisfactory with respect to the additional comp... | [
"Cost-Sensitive",
"Bayesian Optimization",
"Multi-Fidelity HPO",
"PFNs",
"Transfer Learning"
] | null | 29,184 | 2510.21379 | title_snapshot | [
-0.021412178874015808,
-0.00022314413217827678,
-0.007163540925830603,
0.04951826483011246,
0.056215789169073105,
0.0396219901740551,
0.040479592978954315,
-0.011562331579625607,
0.012596228159964085,
-0.03622306510806084,
-0.011422799900174141,
-0.00926928035914898,
-0.06163504719734192,
... |
Novel Exploration via Orthogonality | https://openreview.net/forum?id=yJS1eZSNUv | [
"Andreas Theophilou",
"Özgür Şimşek"
] | Poster | reinforcement_learning | Efficient exploration remains one of the most important open problems in reinforcement learning. Discovering novel states or transitions requires policies that efficiently direct the agent away from the regions of the state space that are already well explored. We introduce Novel Exploration via Orthogonality (NEO), an... | [
"Laplacian",
"Novelty",
"Reinforcement Learning",
"Exploration",
"Eigenvectors",
"Spectral Methods"
] | We use Laplacian representation to improve exploration for reinforcement learning agents. | 29,178 | null | null | [
-0.03849181532859802,
-0.027739834040403366,
0.02891325205564499,
0.04787576198577881,
0.04889220744371414,
0.01062247809022665,
0.024353735148906708,
-0.006604218389838934,
-0.019217785447835922,
-0.06281072646379471,
-0.008614735677838326,
-0.020602980628609657,
-0.06570222228765488,
-0.... |
The Good, the Bad and the Ugly: Meta-Analysis of Watermarks, Transferable Attacks and Adversarial Defenses | https://openreview.net/forum?id=NVDrWBwJTV | [
"Grzegorz Gluch",
"Berkant Turan",
"Sai Ganesh Nagarajan",
"Sebastian Pokutta"
] | Poster | theory | We formalize and analyze the trade-off between backdoor-based watermarks and adversarial defenses, framing it as an interactive protocol between a verifier and a prover. While previous works have primarily focused on this trade-off, our analysis extends it by identifying transferable attacks as a third, counterintuitiv... | [
"Interactive Proof Systems",
"Cryptography",
"Backdoors",
"Game Theory",
"Learning Theory",
"Transferable Attacks",
"Adversarial Robustness"
] | null | 29,164 | 2410.08864 | title_snapshot | [
-0.00954915676265955,
-0.0027254107408225536,
0.0017249038210138679,
0.04232586547732353,
0.03674174100160599,
-0.00445817643776536,
0.0298505499958992,
-0.024170810356736183,
-0.006570352241396904,
-0.04446842148900032,
0.00404771976172924,
0.000050068916607415304,
-0.04198061302304268,
0... |
Improved Algorithms for Overlapping and Robust Clustering of Edge-Colored Hypergraphs: An LP-Based Combinatorial Approach | https://openreview.net/forum?id=F3DrgOZYc6 | [
"Changyeol Lee",
"Yongho Shin",
"Hyung-Chan An"
] | Poster | general_machine_learning | Clustering is a fundamental task in both machine learning and data mining. Among various methods, edge-colored clustering (ECC) has emerged as a useful approach for handling categorical data. Given a hypergraph with (hyper)edges labeled by colors, ECC aims to assign vertex colors to minimize the number of edges where t... | [
"overlapping edge-colored clustering",
"robust edge-colored clustering",
"edge-colored clustering",
"hypergraph clustering",
"primal-dual methods",
"approximation algorithms"
] | This paper presents improved algorithms for overlapping and robust clustering of edge-colored hypergraphs; our algorithms combine the strengths of LP with the efficiency of combinatorial algorithms, efficiently producing high-quality solutions. | 29,157 | 2505.18043 | title_snapshot | [
0.014916172251105309,
-0.006587195210158825,
0.00934670865535736,
0.06634658575057983,
0.052351221442222595,
0.03119027614593506,
-0.008739388547837734,
-0.012193869799375534,
-0.039401739835739136,
-0.04721686616539955,
-0.03083178587257862,
-0.029855700209736824,
-0.08138208836317062,
0.... |
GoalLadder: Incremental Goal Discovery with Vision-Language Models | https://openreview.net/forum?id=BiowiwzQaO | [
"Alexey Zakharov",
"Shimon Whiteson"
] | Poster | reinforcement_learning | Natural language can offer a concise and human-interpretable means of specifying reinforcement learning (RL) tasks. The ability to extract rewards from a language instruction can enable the development of robotic systems that can learn from human guidance; however, it remains a challenging problem, especially in visual... | [
"reinforcement learning",
"vision-language models"
] | Training reinforcement learning agents from a single language instruction using vision-language models. | 29,127 | 2506.16396 | title_snapshot | [
-0.019351663067936897,
-0.01952585205435753,
0.0008987621986307204,
0.011237690225243568,
0.026819022372364998,
0.01259186863899231,
0.03180454298853874,
0.007804885506629944,
-0.04289289191365242,
-0.025205029174685478,
-0.04293089359998703,
0.04514702409505844,
-0.058483242988586426,
-0.... |
CADGrasp: Learning Contact and Collision Aware General Dexterous Grasping in Cluttered Scenes | https://openreview.net/forum?id=CB8jwNE2vV | [
"Jiyao Zhang",
"Zhiyuan Ma",
"Tianhao Wu",
"Zeyuan Chen",
"Hao Dong"
] | Poster | applications | Dexterous grasping in cluttered environments presents substantial challenges due to the high degrees of freedom of dexterous hands, occlusion, and potential collisions arising from diverse object geometries and complex layouts. To address these challenges, we propose CADGrasp, a two-stage algorithm for general dexterou... | [
"Dexterous Hand",
"General Grasping"
] | null | 29,119 | 2601.15039 | title_snapshot | [
-0.023194927722215652,
0.007209736853837967,
-0.0044848802499473095,
0.020418992266058922,
0.018975377082824707,
0.05587774142622948,
-0.016917917877435684,
0.015849120914936066,
-0.04019457846879959,
-0.06050482019782066,
-0.013326494954526424,
-0.032118059694767,
-0.08436792343854904,
-0... |
KLASS: KL-Guided Fast Inference in Masked Diffusion Models | https://openreview.net/forum?id=gOG9Zoyn4R | [
"Seo Hyun Kim",
"Sunwoo Hong",
"Hojung Jung",
"Youngrok Park",
"Se-Young Yun"
] | Spotlight | deep_learning | Masked diffusion models have demonstrated competitive results on various tasks including language generation. However, due to its iterative refinement process, the inference is often bottlenecked by slow and static sampling speed. To overcome this problem, we introduce `KL-Adaptive Stability Sampling' (KLASS), a fast y... | [
"Generative Models",
"Efficient Inference Methods"
] | We propose KLASS, a fast KL-guided sampling method for masked diffusion models that improves accuracy while cutting inference time by over 2x. | 29,103 | 2511.05664 | title_snapshot | [
-0.010106141678988934,
-0.01534708309918642,
0.006160087417811155,
0.05967297777533531,
0.06176680698990822,
0.030737485736608505,
0.03143753111362457,
-0.018340863287448883,
-0.008807222358882427,
-0.05641229450702667,
0.003997195046395063,
-0.028573131188750267,
-0.04740358144044876,
0.0... |
HM3: Hierarchical Multi-Objective Model Merging for Pretrained Models | https://openreview.net/forum?id=JeP0lpusYw | [
"Yu Zhou",
"Xingyu Wu",
"Jibin Wu",
"Liang Feng",
"KC Tan"
] | Spotlight | deep_learning | Model merging is a technique that combines multiple large pretrained models into a single model, enhancing performance and broadening task adaptability without original data or additional training. However, most existing model merging methods focus primarily on exploring the parameter space, merging models with identic... | [
"Large language model",
"model merging",
"multi-objective optimization",
"architecture-level merging"
] | null | 29,077 | 2409.18893 | title_snapshot | [
-0.040831059217453,
0.015840943902730942,
-0.0037798278499394655,
0.04809693619608879,
0.02028195932507515,
0.022017361596226692,
0.01153019443154335,
-0.02592768520116806,
-0.04614310339093208,
-0.06412409245967865,
0.016027551144361496,
0.023893065750598907,
-0.09213187545537949,
-0.0166... |
Router-R1: Teaching LLMs Multi-Round Routing and Aggregation via Reinforcement Learning | https://openreview.net/forum?id=DWf4vroKWJ | [
"Haozhen Zhang",
"Tao Feng",
"Jiaxuan You"
] | Poster | deep_learning | The rapid emergence of diverse large language models (LLMs) has spurred the development of LLM routers that assign user queries to the most suitable model. However, existing LLM routers typically perform a single-round, one-to-one mapping (\textit{i.e.}, assigning each query to a single model in isolation), which limit... | [
"Large Language Models",
"LLM Routers",
"LLM Selection",
"Reinforcement Learning"
] | We propose Router-R1, an RL-based framework that interleaves multi-round reasoning with dynamic LLM selection, supports zero-shot integration of new models, and optimizes performance-cost trade-offs | 29,040 | 2506.09033 | title_snapshot | [
-0.017280833795666695,
-0.04015205428004265,
-0.01185574010014534,
0.028279021382331848,
0.06783580034971237,
-0.008127409964799881,
0.015802603214979172,
0.021499130874872208,
-0.029230616986751556,
-0.03215678036212921,
0.0010691604111343622,
0.025989199057221413,
-0.07260176539421082,
-... |
Structure-Aware Spectral Sparsification via Uniform Edge Sampling | https://openreview.net/forum?id=Z4eFqgYbha | [
"Kaiwen He",
"Petros Drineas",
"Rajiv Khanna"
] | Poster | theory | Spectral clustering is a fundamental method for graph partitioning, but its reliance on eigenvector computation limits scalability to massive graphs. Classical sparsification methods preserve spectral properties by sampling edges proportionally to their effective resistances, but require expensive preprocessing to esti... | [
"Spectral Clustering",
"Graph Sparsification"
] | For Spectral Clustering, Uniform Sampling on the edges works. | 29,032 | 2510.12669 | title_snapshot | [
0.0013320345897227526,
-0.033194396644830704,
0.026841089129447937,
0.04920024797320366,
0.03795435279607773,
0.030123291537165642,
0.0285471323877573,
-0.023917369544506073,
-0.013451331295073032,
-0.08176187425851822,
0.02560131810605526,
-0.039725206792354584,
-0.07180698215961456,
0.00... |
Information-Theoretic Discrete Diffusion | https://openreview.net/forum?id=B2iPEX5A9c | [
"Moongyu Jeon",
"Sangwoo Shin",
"Dongjae Jeon",
"Albert No"
] | Poster | theory | We present an information-theoretic framework for discrete diffusion models
that yields principled estimators of log-likelihood using score-matching losses.
Inspired by the I-MMSE identity for the Gaussian setup, we derive analogous results for the discrete setting.
Specifically, we introduce the Information–Minimum D... | [
"Discrete Diffusion Models",
"Information Theory",
"Score Matching",
"Denoising Score Entropy (DSE)",
"Denoising Cross-Entropy (DCE)"
] | We derive information-theoretic identities for discrete diffusion, revealing score-based losses as exact mutual information decompositions and enabling principled log-likelihood estimation. | 29,019 | 2510.24088 | title_snapshot | [
-0.029163071885704994,
0.01928512379527092,
0.0005815858021378517,
0.030300313606858253,
0.03904833644628525,
0.05054179206490517,
0.041815947741270065,
-0.006145056802779436,
-0.015026122331619263,
-0.057173602283000946,
0.018239693716168404,
-0.0056561739183962345,
-0.0368783138692379,
-... |
From Euler to AI: Unifying Formulas for Mathematical Constants | https://openreview.net/forum?id=cNqMAmpZh4 | [
"Tomer Raz",
"Michael Shalyt",
"Elyasheev Leibtag",
"Rotem Kalisch",
"Shachar Weinbaum",
"Yaron Hadad",
"Ido Kaminer"
] | Poster | machine_learning_for_sciences | The constant $\large \pi$ has fascinated scholars throughout the centuries, inspiring numerous formulas for its evaluation, such as infinite sums and continued fractions. Despite their individual significance, many of the underlying connections among formulas remain unknown, missing unifying theories that could unveil ... | [
"AI for Science",
"AI for Math",
"LLM-Tool Integration",
"Mathematical Constants",
"Continued Fractions",
"Recurrences",
"Number Theory",
"Pi"
] | A general framework for unifying mathematical knowledge, clustering it to prove previously unknown equivalences across scientific literature—demonstrated by connecting historic and modern formulas for 𝜋. | 29,012 | 2502.17533 | title_snapshot | [
-0.018244849517941475,
0.00993821956217289,
-0.0005969606572762132,
-0.007674131076782942,
0.05666767433285713,
0.009102207608520985,
0.016413133591413498,
-0.005704181734472513,
-0.030842024832963943,
-0.014692654833197594,
-0.008615311235189438,
-0.006828350480645895,
-0.0637950748205185,
... |
CVGL: Causal Learning and Geometric Topology | https://openreview.net/forum?id=1CqEAuRzHc | [
"Songsong Ouyang",
"Yingying Zhu"
] | Poster | deep_learning | Cross-view geo-localization (CVGL) aims to estimate the geographic location of a street image by matching it with a corresponding aerial image. This is critical for autonomous navigation and mapping in complex real-world scenarios. However, the task remains challenging due to significant viewpoint differences and the i... | [
"cross-view",
"casual learning",
"BEV"
] | null | 29,010 | 2603.12551 | title_snapshot | [
0.025838611647486687,
-0.014672178775072098,
0.049155283719301224,
0.03178103640675545,
0.006703269202262163,
0.022523710504174232,
0.04050053283572197,
0.04933663457632065,
0.0016909800469875336,
-0.06662575900554657,
-0.04135991632938385,
-0.015310884453356266,
-0.07474349439144135,
0.00... |
Q-Palette: Fractional-Bit Quantizers Toward Optimal Bit Allocation for Efficient LLM Deployment | https://openreview.net/forum?id=l4F50jpiVH | [
"Deokjae Lee",
"Hyun Oh Song"
] | Poster | deep_learning | We study weight-only post-training quantization (PTQ), which quantizes the weights of a large language model (LLM) without retraining, using little or no calibration data. Weight-only PTQ is crucial for reducing the memory footprint and latency of LLM inference, especially in memory-bound, small-batch inference scenari... | [
"LLM quantization",
"Post-training quantization",
"Mixed scheme quantization",
"Data-free quantization"
] | We develop Q-Palette, a quantizer suite with efficient inference CUDA kernels and wide fractional-bit support, enabling mixed-scheme quantization that achieves ~36% faster LLM decoding than NormalFloat while improving accuracy. | 28,978 | 2509.20214 | title_snapshot | [
-0.020840730518102646,
-0.03644438087940216,
-0.021218758076429367,
0.032333262264728546,
0.035752907395362854,
0.04980524629354477,
-0.007585585117340088,
0.011041316203773022,
-0.04620474949479103,
-0.016247522085905075,
-0.05048529803752899,
0.0036172778345644474,
-0.07325447350740433,
... |
Neural Evolution Strategy for Black-box Pareto Set Learning | https://openreview.net/forum?id=pQ0D0vdjJv | [
"Chengyu LU",
"Zhenhua Li",
"Xi Lin",
"Ji Cheng",
"Qingfu Zhang"
] | Poster | optimization | Multi-objective optimization problems (MOPs) are prevalent in numerous real-world applications. Recently, Pareto Set Learning (PSL) has emerged as a powerful paradigm for solving MOPs. PSL can produce a neural network for modeling the set of all Pareto optimal solutions. However, applying PSL to black-box objectives, p... | [
"evolution strategy",
"multi-objective optimization",
"black-box optimization",
"pareto set learning"
] | null | 28,972 | null | null | [
-0.05196322128176689,
-0.0059552183374762535,
-0.010403431951999664,
0.007193692494183779,
0.03820260241627693,
0.07516112923622131,
-0.0065110232681035995,
0.014233076944947243,
-0.051848024129867554,
-0.03538800776004791,
-0.015971004962921143,
0.028033170849084854,
-0.057605914771556854,
... |
Structured Sparse Transition Matrices to Enable State Tracking in State-Space Models | https://openreview.net/forum?id=RDbuSCWhad | [
"Aleksandar Terzic",
"Nicolas Menet",
"Michael Hersche",
"Thomas Hofmann",
"Abbas Rahimi"
] | Spotlight | deep_learning | Modern state-space models (SSMs) often utilize structured transition matrices
which enable efficient computation but pose restrictions on the model’s expressivity,
as measured in terms of the ability to emulate finite-state automata (FSA). While
unstructured transition matrices are optimal in terms of expressivity, the... | [
"State-Space Models",
"Expressiveness",
"Efficiency",
"Matrix Parametrisation",
"State-Tracking",
"Finite-State Automata"
] | We propose a parametrisation of SSM transition matrices that enables SSMs to track states of arbitrary finite-state automata while keeping the cost of the parallel scan comparable to that of diagonal SSMs. | 28,967 | 2509.22284 | title_snapshot | [
-0.04745500534772873,
-0.028234876692295074,
-0.004053567070513964,
0.030921882018446922,
0.051150787621736526,
0.026540009304881096,
-0.0052400571294128895,
0.03178194537758827,
-0.015499104745686054,
-0.0360916405916214,
0.023312367498874664,
-0.022440079599618912,
-0.0606476366519928,
-... |
Towards Generalizable 3D Human Pose Estimation via Ensembles on Flat Loss Landscapes | https://openreview.net/forum?id=nZ4mFzCZIx | [
"Jumin Han",
"Jun-Hee Kim",
"Seong-Whan Lee"
] | Poster | applications | 3D Human Pose Estimation (HPE) is a fundamental task in the computer vision. Generalization in 3D HPE task is crucial due to the need for robustness across diverse environments and datasets. Existing methods often focus on learning relationships between joints to enhance the generalization capability, but the role of t... | [
"Lifting-based 3D Human Pose Estimation"
] | We propose an adaptive scaling mechanism and an ensemble approach that combines flat-region solutions to enhance 3D HPE generalization. | 28,959 | null | null | [
-0.00002327064794371836,
-0.0033899720292538404,
-0.006282551679760218,
0.005250223912298679,
0.020980121567845345,
0.03065570816397667,
0.027818588539958,
-0.004373867996037006,
-0.028937842696905136,
-0.04057488963007927,
-0.023950207978487015,
-0.007508850656449795,
-0.08942686021327972,
... |
Cross-modal Associations in Vision and Language Models: Revisiting the Bouba-Kiki Effect | https://openreview.net/forum?id=gkcU26BOml | [
"Tom Kouwenhoven",
"Kiana Shahrasbi",
"Tessa Verhoef"
] | Poster | evaluation | Recent advances in multimodal models have raised questions about whether vision-and-language models (VLMs) integrate cross-modal information in ways that reflect human cognition. One well-studied test case in this domain is the bouba-kiki effect, where humans reliably associate pseudowords like ‘bouba’ with round shape... | [
"Cross-modal associations",
"Vision-and-Language Models",
"bouba-kiki effect",
"Cognitive science"
] | We re-evaluate whether vision-and-language models exhibit the human-like bouba-kiki effects, using two methods modelled after human experiments. Compared to humans, VLMs fall short in aligning cross-modal associations with human intuitions. | 28,951 | 2507.10013 | title_snapshot | [
0.004685271997004747,
0.028255220502614975,
-0.0043326192535459995,
0.011012813076376915,
0.015849674120545387,
-0.020105309784412384,
0.05527817830443382,
0.05921605974435806,
-0.03881297633051872,
-0.02206672541797161,
-0.03697076439857483,
0.041194766759872437,
-0.06989947706460953,
-0.... |
How Data Mixing Shapes In-Context Learning: Asymptotic Equivalence for Transformers with MLPs | https://openreview.net/forum?id=615vk8hmeH | [
"Samet Demir",
"Zafer Dogan"
] | Poster | theory | Pretrained Transformers demonstrate remarkable in-context learning (ICL) capabilities, enabling them to adapt to new tasks from demonstrations without parameter updates. However, theoretical studies often rely on simplified architectures (e.g., omitting MLPs), plain data models (e.g., linear regression with isotropic i... | [
"Transformer",
"in-context learning",
"nonlinear MLP",
"data mixing",
"Gaussian universality"
] | null | 28,946 | 2510.25753 | title_snapshot | [
-0.006452944595366716,
-0.011810913681983948,
-0.003501998959109187,
0.029380017891526222,
0.011676923371851444,
0.011671582236886024,
0.01173783652484417,
0.018736470490694046,
-0.03088667057454586,
0.0014371126890182495,
-0.034410521388053894,
0.009533916600048542,
-0.059080690145492554,
... |
Generalized Linear Mode Connectivity for Transformers | https://openreview.net/forum?id=KurYdcCbjv | [
"Alexander Theus",
"Alessandro Cabodi",
"Sotiris Anagnostidis",
"Antonio Orvieto",
"Sidak Pal Singh",
"Valentina Boeva"
] | Oral | deep_learning | Understanding the geometry of neural network loss landscapes is a central question in deep learning, with implications for generalization and optimization. A striking phenomenon is $\textit{linear mode connectivity}$ (LMC), where independently trained models can be connected by low- or zero-barrier paths, despite appea... | [
"Neural Network Merging",
"Linear Mode Connectivity",
"Model Re-basin",
"Parameter Space Geometry",
"Transformer",
"Permutation Invariance",
"Model Fusion"
] | We propose a unified framework for model merging that leverages multiple symmetry classes to enable low- and zero-loss interpolation between independently trained Transformer models, including Vision Transformers and GPT-2. | 28,928 | 2506.22712 | title_snapshot | [
-0.003927552606910467,
-0.005458276718854904,
-0.011281103827059269,
0.024012567475438118,
0.013706850819289684,
0.03652973473072052,
0.02130609191954136,
0.025019096210598946,
-0.020206646993756294,
-0.03906462714076042,
-0.005179911851882935,
-0.014842753298580647,
-0.08848191797733307,
... |
Twilight: Adaptive Attention Sparsity with Hierarchical Top-$p$ Pruning | https://openreview.net/forum?id=Ve693NkzcU | [
"Chaofan Lin",
"Jiaming Tang",
"Shuo Yang",
"Hanshuo Wang",
"Tian Tang",
"Boyu Tian",
"Ion Stoica",
"Song Han",
"Mingyu Gao"
] | Spotlight | deep_learning | Leveraging attention sparsity to accelerate long-context large language models (LLMs) has been of great importance recently. However, most existing sparse attention algorithms use a fixed budget of how many tokens to use in their computations. This simple static decision raises critical issues in real-world deployment ... | [
"Large Language Model",
"Sparse Attention",
"Decode",
"KV Cache"
] | We propose a method which exploit KV cache sparsity efficiently and dynamically through Top-P sampling. | 28,905 | 2502.02770 | title_snapshot | [
0.026844600215554237,
-0.020842786878347397,
-0.00553491897881031,
0.010468961670994759,
0.020006371662020683,
0.012438022531569004,
0.028224028646945953,
0.018900996074080467,
-0.02416415512561798,
-0.025179188698530197,
-0.04760335758328438,
0.027796098962426186,
-0.047942109405994415,
0... |
MASTER: Enhancing Large Language Model via Multi-Agent Simulated Teaching | https://openreview.net/forum?id=5GaDcRVgBw | [
"Liang Yue",
"Yihong Tang",
"Kehai Chen",
"Jie Liu",
"Min Zhang"
] | Poster | deep_learning | Instruction fine-tuning is crucial in NLP tasks, enhancing pretrained models' instruction-following capabilities and task-specific performance. However, obtaining high-quality fine-tuning data for large models is challenging due to data collection difficulties and high production costs. To address this, we propose MAST... | [
"Instruction Fine-Tuning；Data Augmentation；Multi-Agent Systems；Natural Language Processing"
] | We introduce MASTER, a multi-agent framework that enhances LLM instruction data through simulated pedagogical interactions, significantly improving reasoning and generalization. | 28,899 | 2506.02689 | title_snapshot | [
-0.019295310601592064,
-0.03039555810391903,
-0.011135485954582691,
0.06032918021082878,
0.05137964338064194,
0.0010072856675833464,
0.051052313297986984,
0.009578779339790344,
-0.032313719391822815,
-0.0237081628292799,
-0.02366163767874241,
0.056850552558898926,
-0.07461094856262207,
-0.... |
Self-Supervised Learning of Graph Representations for Network Intrusion Detection | https://openreview.net/forum?id=5bu1IOOvf0 | [
"Lorenzo Guerra",
"Thomas Chapuis",
"Guillaume Duc",
"Pavlo Mozharovskyi",
"Van-Tam Nguyen"
] | Poster | applications | Detecting intrusions in network traffic is a challenging task, particularly under limited supervision and constantly evolving attack patterns. While recent works have leveraged graph neural networks for network intrusion detection, they often decouple representation learning from anomaly detection, limiting the utility... | [
"Self-supervised learning",
"Graph neural networks",
"Masked autoencoder",
"Anomaly detection",
"Intrusion detection",
"Network security",
"Representation learning"
] | We propose a self-supervised framework that combines GNNs and a Transformer-based masked autoencoder to detect network intrusions by reconstructing flow representations and flagging high-error patterns as anomalies. | 28,895 | 2509.16625 | title_snapshot | [
-0.00603472488000989,
-0.044682782143354416,
0.0035734856501221657,
0.0690988302230835,
0.03461615741252899,
0.022279154509305954,
0.03829353675246239,
-0.008150974288582802,
0.002857799641788006,
-0.034680649638175964,
0.0414184145629406,
-0.008307508192956448,
-0.07207988202571869,
0.015... |
Kernel conditional tests from learning-theoretic bounds | https://openreview.net/forum?id=hJKDwf32Xu | [
"Pierre-François Massiani",
"Christian Fiedler",
"Lukas Haverbeck",
"Friedrich Solowjow",
"Sebastian Trimpe"
] | Poster | theory | We propose a framework for hypothesis testing on conditional probability distributions, which we then use to construct *statistical tests of functionals of conditional distributions*.
These tests identify the inputs where the functionals differ with high probability, and include tests of conditional moments or two-samp... | [
"kernel methods",
"hypothesis testing",
"statistical learning"
] | We propose tests for general functionals of conditional distributions (including the two-sample test) with finite-sample guarantees and dependent data thanks to generalizations of time-uniform uncertainty bounds for kernel ridge regression. | 28,893 | 2506.03898 | title_snapshot | [
-0.05314992740750313,
-0.012555340304970741,
0.016544833779335022,
0.047049131244421005,
0.053056538105010986,
0.05705368146300316,
0.017273465171456337,
-0.0316990464925766,
-0.015509651042521,
-0.023219073191285133,
-0.0013794155092909932,
0.027405856177210808,
-0.0626513883471489,
-0.00... |
SPOT: Scalable Policy Optimization with Trees for Markov Decision Processes | https://openreview.net/forum?id=OR5WyyTESh | [
"Xuyuan Xiong",
"Pedro Chumpitaz-Flores",
"Kaixun Hua",
"Cheng Hua"
] | Poster | reinforcement_learning | Interpretable reinforcement learning policies are essential for high-stakes decision-making, yet optimizing decision tree policies in Markov Decision Processes (MDPs) remains challenging. We propose SPOT, a novel method for computing decision tree policies, which formulates the optimization problem as a mixed-integer l... | [
"Decision Tree",
"Markov Decision Processes"
] | null | 28,891 | 2510.19241 | title_snapshot | [
-0.06295228004455566,
-0.016371669247746468,
-0.02717854268848896,
0.04374706372618675,
0.04542047902941704,
0.0453462190926075,
0.015765290707349777,
-0.018500983715057373,
-0.0405411534011364,
-0.03391360118985176,
-0.011223589070141315,
0.0010056656319648027,
-0.06948130577802658,
-0.01... |
Thinking in Character: Advancing Role-Playing Agents with Role-Aware Reasoning | https://openreview.net/forum?id=geNdDlzKTG | [
"Yihong Tang",
"Kehai Chen",
"Muyun Yang",
"Zheng-Yu Niu",
"Jing Li",
"Tiejun Zhao",
"Min Zhang"
] | Poster | applications | The advancement of Large Language Models (LLMs) has spurred significant interest in Role-Playing Agents (RPAs) for applications such as emotional companionship and virtual interaction. However, recent RPAs are often built on explicit dialogue data, lacking deep, human-like internal thought processes, resulting in super... | [
"Role-playing",
"dialogue generation"
] | null | 28,885 | 2506.01748 | title_snapshot | [
-0.03033328615128994,
-0.007798085454851389,
0.0028008189983665943,
0.037020329385995865,
0.04045414179563522,
0.014802098274230957,
0.016842028126120567,
-0.006631128024309874,
-0.03636058419942856,
-0.01900254562497139,
-0.0533367358148098,
0.023651737719774246,
-0.04902271926403046,
-0.... |
Cyclic Counterfactuals under Shift–Scale Interventions | https://openreview.net/forum?id=erwwuMhTJX | [
"Saptarshi Saha",
"Dhruv Vansraj Rathore",
"Utpal Garain"
] | Poster | theory | Most counterfactual inference frameworks traditionally assume acyclic structural causal models (SCMs), i.e. directed acyclic graphs (DAGs). However, many real-world systems (e.g. biological systems) contain feedback loops or cyclic dependencies that violate acyclicity. In this work, we study counterfactual inference in... | [
"Counterfactuals",
"Causality",
"cyclic SCM"
] | null | 28,882 | 2510.25005 | title_snapshot | [
0.001992960227653384,
-0.028567038476467133,
-0.03385505825281143,
-0.0107511505484581,
0.04552067071199417,
0.00503392331302166,
0.05306445062160492,
0.0269119031727314,
-0.04229041188955307,
-0.02526707760989666,
0.04564306512475014,
0.02833537571132183,
-0.07066710293292999,
0.021528631... |
SAFEx: Analyzing Vulnerabilities of MoE-Based LLMs via Stable Safety-critical Expert Identification | https://openreview.net/forum?id=VwsXmcMyg5 | [
"ZhengLin Lai",
"Mengyao Liao",
"Bingzhe Wu",
"Dong Xu",
"Zebin Zhao",
"Zhihang Yuan",
"Chao Fan",
"Jianqiang Li"
] | Poster | social_and_economic_aspects_of_machine_learning | Large language models with Mixture-of-Experts (MoE) architectures achieve efficiency and scalability, yet their routing mechanisms introduce safety alignment challenges insufficiently addressed by techniques developed for dense models. In this work, the MoE-specific safety risk of positional vulnerability—that safety-a... | [
"Trustworthy AI"
] | We identified a safety issue in the MoE architecture and designed experiments to demonstrate it. | 28,857 | 2506.17368 | title_snapshot | [
-0.009477694518864155,
-0.00846292544156313,
-0.012448905035853386,
0.040551211684942245,
0.05862952396273613,
0.003636200912296772,
0.035634350031614304,
-0.01070792879909277,
-0.029775524511933327,
-0.04220869019627571,
-0.024801112711429596,
0.04765342175960541,
-0.05183853954076767,
-0... |
A geometric framework for momentum-based optimizers for low-rank training | https://openreview.net/forum?id=cCefuzQrjK | [
"Steffen Schotthöfer",
"Timon Klein",
"Jonas Kusch"
] | Poster | deep_learning | Low-rank pre-training and fine-tuning have recently emerged as promising techniques for reducing the computational and storage costs of large neural networks. Training low-rank parameterizations typically relies on conventional optimizers such as heavy ball momentum methods or Adam. In this work, we identify and analyz... | [
"Low-Rank",
"Compression",
"Finetuning",
"Optimization",
"Manifold"
] | null | 28,854 | 2506.17475 | title_snapshot | [
-0.045687105506658554,
-0.02098814584314823,
0.007231623400002718,
0.043602243065834045,
-0.0033757237251847982,
0.05301631987094879,
0.0076032825745642185,
-0.024389570578932762,
-0.015484390780329704,
-0.0423625223338604,
0.0034616817720234394,
-0.0373939573764801,
-0.04872848093509674,
... |
Atomic Thinking of LLMs: Decoupling and Exploring Mathematical Reasoning Abilities | https://openreview.net/forum?id=iBFfb6bGOz | [
"Jiayi Kuang",
"Haojing Huang",
"Yinghui Li",
"Xinnian Liang",
"Zhikun Xu",
"Yangning Li",
"Xiaoyu Tan",
"Chao Qu",
"Meishan Zhang",
"Ying Shen",
"Philip S. Yu"
] | Poster | evaluation | Large Language Models (LLMs) have demonstrated outstanding performance in mathematical reasoning capabilities. However, we argue that current large-scale reasoning models primarily rely on scaling up training datasets with diverse mathematical problems and long thinking chains, which raises questions about whether LLMs... | [
"Large Language Models",
"Mathematical Reasoning",
"Atomic Thinking"
] | We have decoupled the math atomic capabilities of large language models and explored their interaction relationships in mathematical reasoning tasks. | 28,853 | 2509.25725 | title_snapshot | [
-0.028674662113189697,
0.024663958698511124,
-0.01972324773669243,
0.03183705732226372,
0.08515432476997375,
-0.005115487612783909,
0.011630987748503685,
0.02221156843006611,
-0.033768974244594574,
-0.00044827008969150484,
-0.0031503436621278524,
0.0032842697110027075,
-0.05648462474346161,
... |
Unveiling Concept Attribution in Diffusion Models | https://openreview.net/forum?id=dVIx32Lq7J | [
"Quang H Nguyen",
"Hoang Phan",
"Khoa D Doan"
] | Poster | social_and_economic_aspects_of_machine_learning | Diffusion models have shown remarkable abilities in generating realistic and high-quality images from text prompts. However, a trained model remains largely black-box; little do we know about the roles of its components in exhibiting a concept such as objects or styles. Recent works employ causal tracing to localize kn... | [
"generative models",
"diffusion models",
"interpretability",
"concept erasure"
] | We study how model components store knowledge in diffusion models. | 28,844 | 2412.02542 | title_snapshot | [
0.008436045609414577,
-0.009370137937366962,
-0.02923317439854145,
0.07365167886018753,
0.05572560057044029,
0.002282373607158661,
0.02207878977060318,
0.013677327893674374,
-0.004444802645593882,
-0.05109993368387222,
-0.01644166186451912,
-0.001677052234299481,
-0.024601221084594727,
0.0... |
Brain network science modelling of sparse neural networks enables Transformers and LLMs to perform as fully connected | https://openreview.net/forum?id=OM0Qkq9xtY | [
"Yingtao Zhang",
"Diego Cerretti",
"Jialin Zhao",
"Wenjing Wu",
"Ziheng Liao",
"Umberto Michieli",
"Carlo Vittorio Cannistraci"
] | Poster | deep_learning | This study aims to enlarge our current knowledge on the application of brain-inspired network science principles for training artificial neural networks (ANNs) with sparse connectivity. Dynamic sparse training (DST) emulates the synaptic turnover of real brain networks, reducing the computational demands of training an... | [
"dynamic sparse training",
"network science",
"epitopological Learning",
"efficient training"
] | The proposed brain-inspired CHT Soft Rule with Sigmoid Decay Density (CHTss) achieves comparable even better performance compared to fully connected models across various tasks, enabling high sparsity in Transformers and LLMs. | 28,840 | 2501.19107 | title_snapshot | [
-0.026078466325998306,
-0.015571193769574165,
0.0146563071757555,
0.022601930424571037,
0.02918895334005356,
0.03113992139697075,
0.02666821889579296,
0.028236569836735725,
-0.0594107061624527,
-0.05321725830435753,
0.018004072830080986,
0.002589163603261113,
-0.05026368051767349,
0.012801... |
Uncertainty-Sensitive Privileged Learning | https://openreview.net/forum?id=00Bwl1woOJ | [
"Fan-Ming Luo",
"Lei Yuan",
"Yang Yu"
] | Poster | reinforcement_learning | Privileged learning efficiently tackles high-dimensional, partially observable decision-making problems by first training a privileged policy (PP) on low-dimensional privileged observations, and then deriving a deployment policy (DP) either by imitating the PP or coupling it with an observation encoder. However, since ... | [
"Imitation Gap",
"Reinforcement Learning",
"Privileged Learning",
"Teacher-Student Learning"
] | null | 28,834 | null | null | [
-0.026750996708869934,
-0.039153922349214554,
-0.0015600327169522643,
0.04351929947733879,
0.062292952090501785,
0.016739478334784508,
0.030495522543787956,
-0.027054136618971825,
-0.029820499941706657,
-0.045907340943813324,
-0.02688824012875557,
0.0017379114869982004,
-0.06476414948701859,... |
Joint Modeling of fMRI and EEG Imaging Using Ordinary Differential Equation-Based Hypergraph Neural Networks | https://openreview.net/forum?id=qJLPlZSdkb | [
"YanZhang",
"Yang Gao",
"Min Li"
] | Poster | neuroscience_and_cognitive_science | Fusing multimodal brain imaging has been a hot topic since different modalities of brain imaging can provide complementary information. However, due to the size of simultaneous recorded fMRI-EEG dataset being limited and the substantial discrepancy between hemodynamic responses of fMRI and neural oscillations of EEG, t... | [
"fMRI",
"EEG",
"Multimodal modeling"
] | null | 28,826 | null | null | [
-0.01773781143128872,
0.02424662932753563,
0.012808064930140972,
0.012191634625196457,
0.012870744802057743,
0.02449173294007778,
0.05167720839381218,
0.010338688269257545,
-0.022809162735939026,
-0.06964174658060074,
0.006863127462565899,
-0.0003550268884282559,
-0.05429843068122864,
0.00... |
Improving Target Sound Extraction via Disentangled Codec Representations with Privileged Knowledge Distillation | https://openreview.net/forum?id=rew03VaNUJ | [
"Dail Kim",
"Joon-Hyuk Chang"
] | Poster | applications | Target sound extraction aims to isolate target sound sources from an input mixture using a target clue to identify the sounds of interest. To address the challenge posed by the wide variety of sounds, recent work has introduced privileged knowledge distillation (PKD), which utilizes privileged information (PI) about th... | [
"Target Sound Extraction",
"Privileged Knowledge distillation",
"Disentangled Representation Learning",
"Neural Audio Codec",
"Feature-level Knowledge Distillation"
] | This paper proposes DCKD, a privileged knowledge distillation framework for target sound extraction that regulates the amount and flow of target information via neural codec and disentangled representation learning. | 28,819 | null | null | [
0.008218725211918354,
-0.013470594771206379,
-0.019027525559067726,
0.04237247630953789,
0.05633636936545372,
0.007938344962894917,
0.04629574716091156,
-0.024675073102116585,
-0.015112243592739105,
-0.035660602152347565,
-0.025376003235578537,
0.012238864786922932,
-0.04277532547712326,
0... |
MLEP: Multi-granularity Local Entropy Patterns for Generalized AI-generated Image Detection | https://openreview.net/forum?id=Bsska2ayiy | [
"Lin Yuan",
"Xiaowan Li",
"Yan Zhang",
"Jiawei Zhang",
"Hongbo Li",
"Xinbo Gao"
] | Poster | social_and_economic_aspects_of_machine_learning | Advances in image generation technologies have raised growing concerns about their potential misuse, particularly in producing misinformation and deepfakes. This creates an urgent demand for effective methods to detect AI-generated images (AIGIs). While progress has been made, achieving reliable performance across dive... | [
"AI-generated image detection",
"entropy",
"multi-granularity",
"deepfake detection"
] | This paper proposes a novel AIGI detection method based on Multi-granularity Local Entropy Patterns (MLEP), which captures scale- and location-invariant entropy features to improve accuracy and generalization across diverse generative models. | 28,793 | 2504.13726 | title_judge | [
0.02381320297718048,
-0.03408455103635788,
-0.009614902548491955,
0.02930762618780136,
0.03347955271601677,
0.023418018594384193,
0.01429203525185585,
-0.016277065500617027,
-0.03809821233153343,
-0.06601051241159439,
-0.03714318200945854,
0.003256638068705797,
-0.06773462146520615,
0.0077... |
Reasoning Models Sometimes Output Illegible Chains of Thought | https://openreview.net/forum?id=w1TjXJk846 | [
"Arun Jose"
] | Poster | social_and_economic_aspects_of_machine_learning | Language models trained via outcome-based reinforcement learning (RL) to reason using chain-of-thought (CoT) have shown remarkable performance. Monitoring such a model's CoT may allow us to understand its intentions and detect potential malicious behavior. However, to be effective, this requires that CoTs are legible a... | [
"Reasoning Models Sometimes Output Illegible Chains of Thought"
] | We find that reasoning traces of a RL-trained model often have illegible segments, potentially compromising chain-of-thought monitoring for detecting malicious behavior. | 28,786 | 2510.27338 | title_snapshot | [
0.006798814050853252,
-0.012977835722267628,
-0.022243957966566086,
0.05015222728252411,
0.039901044219732285,
-0.007624698802828789,
0.028129272162914276,
0.02960137650370598,
-0.036625880748033524,
-0.012609247118234634,
-0.034668661653995514,
0.06493112444877625,
-0.057896122336387634,
... |
JanusDNA: A Powerful Bi-directional Hybrid DNA Foundation Model | https://openreview.net/forum?id=9PL1DIIB7e | [
"Qihao Duan",
"Bingding Huang",
"Zhenqiao Song",
"Irina Lehmann",
"Lei Gu",
"Roland Eils",
"Benjamin Wild"
] | Poster | machine_learning_for_sciences | Large language models (LLMs) have revolutionized natural language processing and are increasingly applied to other sequential data types, including genetic sequences. However, adapting LLMs to genetics presents significant challenges. Capturing complex genomic interactions requires modeling long-range global dependenci... | [
"genomics",
"foundation model",
"hybrid architecture",
"learning efficiency"
] | JanusDNA, the first bidirectional DNA foundation model built upon a novel pretraining paradigm, integrating the optimization efficiency of autoregressive modeling with the bidirectional comprehension capability of masked modeling. | 28,779 | 2505.17257 | title_snapshot | [
-0.0013791167875751853,
-0.02600034326314926,
-0.025656644254922867,
0.00003616164030972868,
0.03658641129732132,
0.021986056119203568,
0.02544422261416912,
0.02188948728144169,
0.0019842693582177162,
-0.03213031589984894,
0.03482316434383392,
-0.00976746715605259,
-0.07673951983451843,
0.... |
PubSub-VFL: Towards Efficient Two-Party Split Learning in Heterogeneous Environments via Publisher/Subscriber Architecture | https://openreview.net/forum?id=B5mEYUJi85 | [
"Yi Liu",
"Yang Liu",
"Leqian Zheng",
"Jue Hong",
"Junjie Shi",
"Qingyou Yang",
"Ye Wu",
"Cong Wang"
] | Poster | infrastructure | With the rapid advancement of the digital economy, data collaboration between organizations has become a well-established business model, driving the growth of various industries. However, privacy concerns make direct data sharing impractical. To address this, Two-Party Split Learning (a.k.a. Vertical Federated Learni... | [
"Vertical Federated Learning",
"Publisher/Subscriber Architecture",
"Computational Resource Utilization",
"Asynchronous Mechanism"
] | We propose PubSub-VFL, a novel VFL paradigm with a Publisher/Subscriber architecture optimized for two-party collaborative learning with high computational efficiency. | 28,764 | 2510.12494 | title_snapshot | [
-0.0007952533196657896,
-0.044933170080184937,
0.024434635415673256,
0.03674307093024254,
0.028505241498351097,
0.007917954586446285,
0.03576088324189186,
-0.02879462018609047,
-0.045848555862903595,
-0.01716194860637188,
-0.0001085874464479275,
-0.0316128134727478,
-0.04746098816394806,
0... |
Bohdi: Heterogeneous LLM Fusion with Automatic Data Exploration | https://openreview.net/forum?id=wVxIBvUAlj | [
"Junqi Gao",
"Zhichang Guo",
"Dazhi Zhang",
"Dong Li",
"Runze Liu",
"Pengfei Li",
"Kai Tian",
"Biqing Qi"
] | Poster | applications | Heterogeneous Large Language Model (LLM) fusion integrates the strengths of multiple source LLMs with different architectures into a target LLM with low computational overhead. While promising, existing methods suffer from two major limitations: 1) **reliance on real data from limited domain** for knowledge fusion, pre... | [
"Heterogeneous Model Fusion",
"Large Language Models"
] | null | 28,761 | 2506.15721 | title_snapshot | [
-0.05203660950064659,
-0.006061994470655918,
-0.016765248030424118,
0.018650570884346962,
0.028447741642594337,
0.005188124720007181,
0.043933600187301636,
-0.00961984507739544,
-0.0329999104142189,
0.004616982769221067,
-0.030559442937374115,
0.029275745153427124,
-0.04723895713686943,
0.... |
How Does Sequence Modeling Architecture Influence Base Capabilities of Pre-trained Language Models? Exploring Key Architecture Design Principles to Avoid Base Capabilities Degradation | https://openreview.net/forum?id=vMkJWaa02n | [
"Xin Lu",
"Yanyan Zhao",
"Si Wei",
"Shijin Wang",
"Bing Qin",
"Ting Liu"
] | Poster | deep_learning | Pre-trained language models represented by the Transformer have been proven to possess strong base capabilities, and the representative self-attention mechanism in the Transformer has become a classic in sequence modeling architectures. Different from the work of proposing sequence modeling architecture to improve the ... | [
"Pre-trained Language Models",
"Base Capabilities",
"Sequence Modeling"
] | null | 28,756 | 2505.18522 | title_snapshot | [
-0.027363762259483337,
-0.02041797898709774,
-0.003817289136350155,
0.02826334722340107,
0.05318542197346687,
-0.004321602638810873,
0.03894858434796333,
0.02559499442577362,
0.00000983988957159454,
-0.00980566255748272,
-0.011022514663636684,
0.03099709376692772,
-0.040584225207567215,
0.... |
Iterative Missing Data Imputation with Model Form Adaptation and Non-Missing Feature Supervision | https://openreview.net/forum?id=L84DdFuvwV | [
"Hao Wang",
"zhengnan li",
"Zhichao Chen",
"Xu Chen",
"Shuting He",
"Guangyi Liu",
"Haoxuan Li",
"Zhouchen Lin"
] | Poster | applications | Iterative imputation is a prevalent method for missing data imputation, where each feature is imputed iteratively by treating it as a target variable estimated from all other features. However, iterative imputation method suffers from two principal limitations:
(1) it imposes a single parametric model form to impute a... | [
"missing data imputation",
"missing data completion",
"missing value imputation",
"kernel",
"ridge regression",
"non-missing feature"
] | null | 28,751 | null | null | [
-0.03389904275536537,
-0.046540118753910065,
-0.0089633259922266,
0.07092051208019257,
0.04696127027273178,
0.06816267222166061,
0.034696150571107864,
-0.017147032544016838,
-0.037654247134923935,
-0.03373385965824127,
-0.04648739844560623,
0.005969339050352573,
-0.035928770899772644,
0.02... |
Reinforcement Learning with Backtracking Feedback | https://openreview.net/forum?id=14B5d6NEaH | [
"Bilgehan Sel",
"Vaishakh Keshava",
"Phillip Wallis",
"Lukas Rutishauser",
"Ming Jin",
"Dingcheng Li"
] | Poster | deep_learning | Addressing the critical need for robust safety in Large Language Models (LLMs), particularly against adversarial attacks and in-distribution errors, we introduce Reinforcement Learning with Backtracking Feedback (RLBF). This framework advances upon prior methods, such as BSAFE, by primarily leveraging a Reinforcement L... | [
"large language models",
"safety alignment",
"reinforcement learning"
] | A backtracking method that reverts to safer points during generation, reducing safety violations | 28,718 | 2602.08377 | title_snapshot | [
-0.017970217391848564,
-0.026220237836241722,
-0.006019747816026211,
0.03249557688832283,
0.04051907733082771,
0.02776261419057846,
0.04521984979510307,
0.02264142595231533,
-0.05124546214938164,
-0.022411897778511047,
-0.02170500159263611,
0.0506480447947979,
-0.05615374445915222,
-0.0196... |
Exploring Landscapes for Better Minima along Valleys | https://openreview.net/forum?id=XxRKqFsvoK | [
"Tong Zhao",
"Jiacheng Li",
"Yuanchang Zhou",
"Guangming Tan",
"Weile Jia"
] | Poster | optimization | Finding lower and better-generalizing minima is crucial for deep learning. However, most existing optimizers stop searching the parameter space once they reach a local minimum. Given the complex geometric properties of the loss landscape, it is difficult to guarantee that such a point is the lowest or provides the best... | [
"optimization",
"landscape",
"exploration",
"local minimum",
"convergence",
"exponential moving average"
] | null | 28,714 | 2510.27153 | title_snapshot | [
-0.02677713893353939,
-0.020288093015551567,
0.02001846209168434,
0.005606680177152157,
0.03531515225768089,
0.052924126386642456,
0.029320128262043,
0.0038594724610447884,
-0.03147746995091438,
-0.06002947688102722,
-0.009269379079341888,
-0.0072111873887479305,
-0.04885444417595863,
0.00... |
Safe RLHF-V: Safe Reinforcement Learning from Multi-modal Human Feedback | https://openreview.net/forum?id=OIH3T5ZPBW | [
"Jiaming Ji",
"Xinyu Chen",
"Rui Pan",
"Han Zhu",
"Jiahao Li",
"Donghai Hong",
"Boyuan Chen",
"Jiayi Zhou",
"Kaile Wang",
"Juntao Dai",
"Chi-Min Chan",
"Sirui Han",
"Yike Guo",
"Yaodong Yang"
] | Poster | social_and_economic_aspects_of_machine_learning | Multimodal large language models (MLLMs) are essential for building general-purpose AI assistants; however, they pose increasing safety risks. How can we ensure safety alignment of MLLMs to prevent undesired behaviors? Going further, it is critical to explore how to fine-tune MLLMs to preserve capabilities while meetin... | [
"AI Safety",
"AI Alignment"
] | Safe RLHF-V, the multimodal safety alignment framework. | 28,710 | 2503.17682 | title_snapshot | [
0.008131972514092922,
0.007793562952429056,
0.0005029692547395825,
0.041322410106658936,
0.031003499403595924,
0.004794177133589983,
0.03637072071433067,
0.003852640511468053,
-0.032458722591400146,
-0.025012031197547913,
-0.021393895149230957,
0.043028172105550766,
-0.08318450301885605,
-... |
World-aware Planning Narratives Enhance Large Vision-Language Model Planner | https://openreview.net/forum?id=fggSyPPk0K | [
"Junhao Shi",
"Zhaoye Fei",
"Siyin Wang",
"Qipeng Guo",
"Jingjing Gong",
"Xipeng Qiu"
] | Poster | reinforcement_learning | Large Vision-Language Models (LVLMs) show promise for embodied planning tasks but struggle with complex scenarios involving unfamiliar environments and multi-step goals.
Current approaches rely on environment-agnostic imitation learning that disconnects instructions from environmental contexts, causing models to strug... | [
"planning",
"embodied",
"LVLMs"
] | null | 28,707 | 2506.21230 | title_snapshot | [
0.02328207716345787,
-0.005036559421569109,
0.01014516782015562,
0.015256958082318306,
0.03503720834851265,
0.005412898492068052,
0.015337239019572735,
0.01192901935428381,
-0.026753000915050507,
-0.010732471011579037,
-0.057714708149433136,
0.026363220065832138,
-0.06183721870183945,
-0.0... |
An Analysis of Causal Effect Estimation using Outcome Invariant Data Augmentation | https://openreview.net/forum?id=C1LVIInfZO | [
"UZAIR AKBAR",
"Niki Kilbertus",
"Hao Shen",
"Krikamol Muandet",
"Bo Dai"
] | Spotlight | probabilistic_methods | The technique of data augmentation (DA) is often used in machine learning for regularization purposes to better generalize under i.i.d. settings. In this work, we present a unifying framework with topics in causal inference to make a case for the use of DA beyond just the i.i.d. setting, but for generalization across i... | [
"Causal Inference",
"Data Augmentation",
"Instrumental Variables",
"Out-of-distribution Generalization",
"Causal Regularization"
] | We show the effectiveness of data-augmentation for mitigating bias due to unobserved confounding, and this motivates the proposal of our novel method for the same. | 28,683 | 2510.25128 | title_snapshot | [
-0.001684293383732438,
-0.018419621512293816,
-0.020353233441710472,
0.043711937963962555,
0.037662748247385025,
0.03279373422265053,
0.06347453594207764,
-0.012104910798370838,
-0.034827347844839096,
-0.04088051989674568,
-0.02489338256418705,
-0.0037197626661509275,
-0.07793533802032471,
... |
Embodied Cognition Augmented End2End Autonomous Driving | https://openreview.net/forum?id=0MXUkBmm09 | [
"Ling Niu",
"Xiaoji Zheng",
"han wang",
"Ziyuan Yang",
"Chen Zheng",
"Bokui Chen",
"Jiangtao Gong"
] | Poster | deep_learning | In recent years, vision-based end-to-end autonomous driving has emerged as a new paradigm. However, popular end-to-end approaches typically rely on visual feature extraction networks trained under label supervision. This limited supervision framework restricts the generality and applicability of driving models. In this... | [
"End-to-End Autonomous Driving",
"Cognitive-Enhanced AI Algorithms"
] | Our work the first to incorporate human driving cognition to enhance end-to-end autonomous driving models, yielding significant findings. | 28,666 | 2511.01334 | title_snapshot | [
0.013109122402966022,
0.005318008363246918,
0.010284286923706532,
0.019813023507595062,
0.013805319555103779,
-0.00347789004445076,
0.0220257006585598,
0.012694436125457287,
-0.024989204481244087,
-0.04165772721171379,
-0.03447180241346359,
-0.00874969083815813,
-0.04548395425081253,
-0.02... |
Feature-aware Modulation for Learning from Temporal Tabular Data | https://openreview.net/forum?id=88MXvVn5dl | [
"Haorun Cai",
"Han-Jia Ye"
] | Poster | deep_learning | While tabular machine learning has achieved remarkable success, temporal distribution shifts pose significant challenges in real-world deployment, as the relationships between features and labels continuously evolve. Static models assume fixed mappings to ensure generalization, whereas adaptive models may overfit to tr... | [
"tabular deep learning",
"distribution shift",
"temporal shift",
"machine learning"
] | null | 28,660 | 2512.03678 | title_snapshot | [
-0.027918558567762375,
-0.022952809929847717,
-0.005360777955502272,
0.0226063784211874,
0.04730524495244026,
0.012552480213344097,
0.02601015754044056,
0.011746062897145748,
-0.029478244483470917,
-0.020481059327721596,
-0.021653931587934494,
0.01032445952296257,
-0.05411664769053459,
0.0... |
Seemingly Redundant Modules Enhance Robust Odor Learning in Fruit Flies | https://openreview.net/forum?id=d6WUTRJqP3 | [
"HaiYang Li",
"Liao Yu",
"Qiang Yu",
"Yunliang Zang"
] | Poster | neuroscience_and_cognitive_science | Biological circuits have evolved to incorporate multiple modules that perform similar functions. In the fly olfactory circuit, both lateral inhibition (LI) and neuronal spike frequency adaptation (SFA) are thought to enhance pattern separation for odor learning. However, it remains unclear whether these mechanisms play... | [
"olfactory representations; lateral inhibition；spike frequency adaptation;pattern classification；spiking neural networks"
] | null | 28,647 | 2510.21315 | title_snapshot | [
-0.00045336096081882715,
0.03503010794520378,
-0.02542884647846222,
0.005083353724330664,
0.03972636163234711,
0.0031429920345544815,
0.03907417505979538,
0.008002142421901226,
-0.027449579909443855,
-0.0563650019466877,
0.02656022645533085,
-0.009238634258508682,
-0.07949750870466232,
0.0... |
Brain-tuning Improves Generalizability and Efficiency of Brain Alignment in Speech Models | https://openreview.net/forum?id=4jgsUhWWaF | [
"Omer Moussa",
"Mariya Toneva"
] | Poster | neuroscience_and_cognitive_science | Pretrained language models are remarkably effective in aligning with human brain responses elicited by natural language stimuli, positioning them as promising model organisms for studying language processing in the brain. However, existing approaches for both estimating and improving this brain alignment are participan... | [
"fMRI",
"Alignment",
"Brain Alignment",
"Cognitive Neuroscience",
"Encoding Models",
"Speech Models"
] | Fine-tuning speech models jointly with brain responses from multiple participants improves brain alignment (even on novel datasets) and decreases the amount of fMRI data needed to predict new participants. | 28,632 | 2510.21520 | title_snapshot | [
-0.008057652041316032,
-0.002047710819169879,
-0.0100777642801404,
0.025381149724125862,
0.030011070892214775,
0.0384693369269371,
0.056912701576948166,
0.029248971492052078,
-0.02621774934232235,
-0.03708610683679581,
-0.026071971282362938,
0.016177823767066002,
-0.07548335194587708,
-0.0... |
Multi-Agent Imitation by Learning and Sampling from Factorized Soft Q-Function | https://openreview.net/forum?id=RbkHARGCcH | [
"Yi-Chen Li",
"Zhongxiang Ling",
"Tao Jiang",
"Fuxiang Zhang",
"Pengyuan Wang",
"Lei Yuan",
"Zongzhang Zhang",
"Yang Yu"
] | Poster | reinforcement_learning | Learning from multi-agent expert demonstrations, known as Multi-Agent Imitation Learning (MAIL), provides a promising approach to sequential decision-making. However, existing MAIL methods including Behavior Cloning (BC) and Adversarial Imitation Learning (AIL) face significant challenges: BC suffers from the compoundi... | [
"Multi-Agent Imitation Learning",
"Energy-Based Models",
"Inverse Reinforcement Learning"
] | null | 28,624 | null | null | [
-0.019825845956802368,
-0.038257986307144165,
0.0049230181612074375,
0.040106937289237976,
0.03602264076471329,
0.02510405331850052,
0.021019158884882927,
0.003578574163839221,
-0.026776917278766632,
-0.040703900158405304,
-0.0018523021135479212,
0.004731255117803812,
-0.10383635014295578,
... |
Root Cause Analysis of Outliers with Missing Structural Knowledge | https://openreview.net/forum?id=7Nxq4RQApu | [
"William Roy Orchard",
"Nastaran Okati",
"Sergio Hernan Garrido Mejia",
"Patrick Blöbaum",
"Dominik Janzing"
] | Poster | probabilistic_methods | The goal of Root Cause Analysis (RCA) is to explain why an anomaly occurred by identifying where the fault originated. Several recent works model the anomalous event as resulting from a change in the causal mechanism at the root cause, i.e., as a soft intervention. RCA is then the task of identifying which causal mecha... | [
"root cause analysis",
"causality",
"contribution analysis",
"actual causation",
"outliers",
"anomalies"
] | null | 28,604 | 2406.05014 | title_snapshot | [
-0.014753956347703934,
-0.01865684799849987,
-0.018353521823883057,
0.049838002771139145,
0.059766095131635666,
0.03557566553354263,
0.03162913769483566,
-0.004069786053150892,
-0.02989119477570057,
-0.03390343114733696,
-0.000408458843594417,
0.025130679830908775,
-0.04641399160027504,
0.... |
Deciphering the Extremes: A Novel Approach for Pathological Long-tailed Recognition in Scientific Discovery | https://openreview.net/forum?id=E16vULI6AF | [
"Zhe Zhao",
"HaiBin Wen",
"Xianfu Liu",
"Rui Mao",
"Pengkun Wang",
"Liheng Yu",
"Linjiang Chen",
"Bo An",
"Qingfu Zhang",
"Yang Wang"
] | Spotlight | general_machine_learning | Scientific discovery across diverse fields increasingly grapples with datasets exhibiting pathological long-tailed distributions: a few common phenomena overshadow a multitude of rare yet scientifically critical instances. Unlike standard benchmarks, these scientific datasets often feature extreme imbalance coupled wit... | [
"Long-tailed learning",
"Imbalanced datasets"
] | null | 28,592 | null | null | [
-0.014611810445785522,
-0.03515714034438133,
-0.0027724364772439003,
0.007176895160228014,
0.051922205835580826,
-0.02142626792192459,
0.002755175344645977,
-0.008580896072089672,
0.005032943096011877,
-0.02836497314274311,
-0.005128296557813883,
0.005393065046519041,
-0.05581940338015556,
... |
Causal Mixture Models: Characterization and Discovery | https://openreview.net/forum?id=aI3d897dgV | [
"Sarah Mameche",
"Janis Kalofolias",
"Jilles Vreeken"
] | Poster | general_machine_learning | Real-world datasets are often a combination of unobserved subpopulations that follow distinct causal generating processes. In an observational study, for example, participants may fall into unknown groups that either (a) respond effectively to a drug, or (b) show no response due to drug resistance. Not accounting for s... | [
"causal discovery",
"mixture modelling"
] | Given a mixture of samples from unobserved subpopulations with distinct underlying causal mechanisms, we give results on identification and discovery of causal graph with latent mixing variables. | 28,586 | null | null | [
0.005950316786766052,
-0.030033709481358528,
-0.022305388003587723,
0.031628530472517014,
0.033930789679288864,
0.016102639958262444,
0.04758020490407944,
0.004040095955133438,
-0.01726909913122654,
-0.035597946494817734,
-0.0007653778302483261,
0.0077363126911222935,
-0.0476636178791523,
... |
From Dormant to Deleted: Tamper-Resistant Unlearning Through Weight-Space Regularization | https://openreview.net/forum?id=Zrqn7ZshXG | [
"Shoaib Ahmed Siddiqui",
"Adrian Weller",
"David Krueger",
"Gintare Karolina Dziugaite",
"Michael Curtis Mozer",
"Eleni Triantafillou"
] | Poster | deep_learning | Recent unlearning methods for LLMs are vulnerable to relearning attacks: knowledge believed-to-be-unlearned re-emerges by fine-tuning on a small set of (even seemingly-unrelated) examples. We study this phenomenon in a controlled setting for example-level unlearning in vision classifiers. We make the surprising discove... | [
"Unlearning",
"tamper-resistance",
"relearning attacks",
"weight-space analysis"
] | We highlight the susceptibility of existing unlearning methods to relearning attacks and analyze the characteristics of robust methods by leveraging the weight-space perspective. | 28,577 | 2505.22310 | title_snapshot | [
0.006250745616853237,
-0.027736233547329903,
0.012443975545465946,
0.04281747341156006,
0.07531224936246872,
-0.01895245909690857,
0.02880174294114113,
-0.0012990464456379414,
-0.04069371521472931,
-0.022028857842087746,
-0.0019070310518145561,
0.027787944301962852,
-0.06374021619558334,
-... |
TV-Rec: Time-Variant Convolutional Filter for Sequential Recommendation | https://openreview.net/forum?id=cWEssTIwG5 | [
"Yehjin Shin",
"Jeongwhan Choi",
"Seojin Kim",
"Noseong Park"
] | Poster | deep_learning | Recently, convolutional filters have been increasingly adopted in sequential recommendation for their ability to capture local sequential patterns. However, most of these models complement convolutional filters with self-attention. This is because convolutional filters alone, generally fixed filters, struggle to captur... | [
"Sequential Recommendation",
"Graph Signal Processing"
] | TV-Rec uses time-variant convolutional filters to model complex user behavior without self-attention, outperforming state-of-the-art baseline models across 6 datasets. | 28,568 | 2510.25259 | title_snapshot | [
0.03460843488574028,
-0.03013581968843937,
0.028364429250359535,
0.01299052219837904,
0.04314076155424118,
0.007378997281193733,
0.021428894251585007,
0.03980420157313347,
0.015494891442358494,
-0.04150833189487457,
-0.00341777759604156,
0.02246013842523098,
-0.053726885467767715,
0.009099... |
AutoSciDACT: Automated Scientific Discovery through Contrastive Embedding and Hypothesis Testing | https://openreview.net/forum?id=vKyiv67VWa | [
"Samuel Bright-Thonney",
"Christina Reissel",
"Gaia Grosso",
"Nathaniel S. Woodward",
"Katya Govorkova",
"Andrzej Novak",
"Sang Eon Park",
"Eric A. Moreno",
"Philip Harris"
] | Poster | machine_learning_for_sciences | Novelty detection in large scientific datasets faces two key challenges: the noisy and high-dimensional nature of experimental data,
and the necessity of making *statistically robust* statements about any observed outliers. While there is a wealth of literature on anomaly detection via dimensionality reduction, most me... | [
"anomaly detection",
"contrastive learning",
"scientific discovery",
"hypothesis testing",
"automated discovery"
] | null | 28,560 | 2510.21935 | title_snapshot | [
0.006278974935412407,
-0.0582871176302433,
-0.008685210719704628,
0.061328716576099396,
0.05026024580001831,
-0.020811384543776512,
0.019591476768255234,
-0.0440969355404377,
-0.0415780246257782,
-0.039272844791412354,
-0.02169743739068508,
0.005881702061742544,
-0.05335305258631706,
0.019... |
Multivariate Latent Recalibration for Conditional Normalizing Flows | https://openreview.net/forum?id=nO8ShqG2ci | [
"Victor Dheur",
"Souhaib Ben Taieb"
] | Poster | probabilistic_methods | A reliable estimate of the full conditional distribution of a multivariate response given a set of covariates is essential in many decision-making applications. However, misspecified or miscalibrated models can lead to poor approximations of the joint distribution, resulting in unreliable predictions and suboptimal dec... | [
"Uncertainty Quantification",
"Model Calibration",
"Multi-response regression",
"Model Recalibration",
"Generative Models"
] | Latent recalibration learns a radial transform that calibrates normalizing flows, preserving an explicit PDF and improving NLL. | 28,554 | 2505.16636 | title_snapshot | [
0.021738387644290924,
-0.03423289954662323,
-0.007605960126966238,
0.04994123429059982,
0.03519774600863457,
0.049366045743227005,
0.01770605891942978,
-0.0046003288589417934,
-0.013828766532242298,
-0.053423039615154266,
-0.011972053907811642,
-0.0007512536249123514,
-0.05578267574310303,
... |
Teaching Language Models to Reason with Tools | https://openreview.net/forum?id=kRZVz1qEqa | [
"Chengpeng Li",
"Zhengyang Tang",
"Ziniu Li",
"Mingfeng Xue",
"Keqin Bao",
"Tian Ding",
"Ruoyu Sun",
"Benyou Wang",
"Xiang Wang",
"Junyang Lin",
"Dayiheng Liu"
] | Poster | applications | Large reasoning models (LRMs) like OpenAI-o1 have shown impressive capabilities in natural language reasoning. However, these models frequently demonstrate inefficiencies or inaccuracies when tackling complex mathematical operations. While integrating computational tools such as Code Interpreters (CIs) offers a promisi... | [
"Tool-integrated Reasoning",
"Large Reasoning Model",
"Long Chain-of-Thought"
] | This paper introduce CoRT, a post-training framework for teaching large reasoning LLMs to leverage CI effectively and efficiently. | 28,540 | 2510.20342 | title_snapshot | [
-0.022685682401061058,
-0.02384086139500141,
-0.05244951322674751,
0.07040974497795105,
0.04552407190203667,
0.025324933230876923,
0.011884677223861217,
0.012748301960527897,
0.007854243740439415,
-0.006519207265228033,
-0.017063874751329422,
0.046510759741067886,
-0.06037634611129761,
-0.... |
Generative Modeling of Full-Atom Protein Conformations using Latent Diffusion on Graph Embeddings | https://openreview.net/forum?id=JPjMXgQQxk | [
"Aditya Sengar",
"Ali Hariri",
"Daniel Probst",
"PATRICK BARTH",
"Pierre Vandergheynst"
] | Poster | machine_learning_for_sciences | Generating diverse, all‐atom conformational ensembles of dynamic proteins such as G‐protein‐coupled receptors (GPCRs) is critical for understanding their function, yet most generative models simplify atomic detail or ignore conformational diversity altogether. We present latent diffusion for full protein generation (LD... | [
"Latent Diffusion",
"Graph Neural Networks",
"Protein Structure Generation",
"All-atom modeling",
"Molecular Dynamics"
] | null | 28,539 | 2506.17064 | title_snapshot | [
-0.013066458515822887,
-0.0031707820016890764,
-0.025794275104999542,
0.05856922268867493,
0.034954849630594254,
0.011008083820343018,
0.014695432037115097,
-0.006062107626348734,
0.02630673348903656,
-0.044911839067935944,
0.01300216093659401,
-0.021673379465937614,
-0.0539889857172966,
0... |
Price of Parsimony: Complexity of Fourier Sparsity Testing | https://openreview.net/forum?id=7bCPXHq8xV | [
"Arijit Ghosh",
"Manmatha Roy"
] | Poster | theory | A function \( f : \mathbb{F}_2^n \to \mathbb{R} \) is said to be \( s \)-Fourier sparse if its Fourier expansion contains at most \( s \) nonzero coefficients. In general, the existence of a sparse representation in the Fourier basis serves as a key enabler for the design of efficient learning algorithms. However, most... | [
"Fourier Sparsity",
"Boolean Function",
"Fourier Analysis",
"Computational Learning Theory",
"Property Testing"
] | The authors design a query-efficient algorithm that, given oracle access to a real valued function over Boolean cube, estimates its $\ell_2^2$ distance to the nearest $k$-Fourier sparse real valued function defined over Boolean Cube. | 28,508 | null | null | [
-0.04381278157234192,
-0.00987785030156374,
0.0141722671687603,
0.04122627153992653,
0.0364394448697567,
0.0335218720138073,
0.01658637635409832,
-0.009810938499867916,
-0.03840373829007149,
-0.04273205250501633,
0.03018801286816597,
0.0016770937945693731,
-0.05114484578371048,
0.008028802... |
CoCoA: A Minimum Bayes Risk Framework Bridging Confidence and Consistency for Uncertainty Quantification in LLMs | https://openreview.net/forum?id=H1NGlLNaVC | [
"Roman Vashurin",
"Maiya Goloburda",
"Albina Ilina",
"Aleksandr Rubashevskii",
"Preslav Nakov",
"Artem Shelmanov",
"Maxim Panov"
] | Poster | deep_learning | Uncertainty quantification for Large Language Models (LLMs) encompasses a diverse range of approaches, with two major families being particularly prominent: (i) information-based, which estimate model confidence from token-level probabilities, and (ii) consistency-based, which assess the semantic agreement among multip... | [
"LLM",
"Large Language Model",
"Uncertainty Quantification",
"Minimum Bayes Risk"
] | A new method of uncertainty quantification for LLMs based on minimum Bayes risk framework combines model confidence with observed consistency. | 28,493 | 2502.04964 | title_judge | [
-0.008427712135016918,
0.011825582012534142,
-0.029665352776646614,
0.04584881290793419,
0.05314568802714348,
0.04223482310771942,
0.023203633725643158,
0.017267510294914246,
-0.06470569968223572,
-0.01835741475224495,
-0.03194211423397064,
0.041996415704488754,
-0.04240045323967934,
-0.00... |
OpenCUA: Open Foundations for Computer-Use Agents | https://openreview.net/forum?id=6iRZvJiC9Q | [
"Xinyuan Wang",
"Bowen Wang",
"Dunjie Lu",
"Junlin Yang",
"Tianbao Xie",
"Junli Wang",
"Jiaqi Deng",
"Xiaole Guo",
"Yiheng Xu",
"Chen Henry Wu",
"Zhennan Shen",
"Zhuokai Li",
"Ryan Li",
"Xiaochuan Li",
"Junda Chen",
"Zheng Boyuan",
"LI PEIHANG",
"Fangyu Lei",
"Ruisheng Cao",
"Y... | Spotlight | applications | Vision-language models have demonstrated impressive capabilities as computer-use agents (CUAs) capable of automating diverse computer tasks. As their commercial potential grows, critical details of the most capable CUA systems remain closed. As these agents will increasingly mediate digital interactions and execute con... | [
"Computer-Use Agent",
"Visual Language Model",
"Planning",
"Reasoning",
"Scaling",
"Dataset",
"Evaluation"
] | null | 28,488 | 2508.09123 | title_snapshot | [
0.00017850700533017516,
-0.028569960966706276,
-0.021673331037163734,
0.03165462240576744,
0.03618639335036278,
0.015468425117433071,
0.027033181861042976,
0.05325734615325928,
-0.020091434940695763,
-0.02346101775765419,
-0.008255112916231155,
0.0023415579926222563,
-0.11057180166244507,
... |
On the Global Optimality of Policy Gradient Methods in General Utility Reinforcement Learning | https://openreview.net/forum?id=aq9Nc5NvNc | [
"Anas Barakat",
"Souradip Chakraborty",
"Peihong Yu",
"Pratap Tokekar",
"Amrit Singh Bedi"
] | Poster | reinforcement_learning | Reinforcement learning with general utilities (RLGU) offers a unifying framework to capture several problems beyond standard expected returns, including imitation learning, pure exploration, and safe RL. Despite recent fundamental advances in the theoretical analysis of policy gradient (PG) methods for standard RL and ... | [
"policy gradient methods",
"reinforcement learning with general utilities"
] | null | 28,478 | 2410.04108 | title_snapshot | [
-0.043640729039907455,
-0.02812390774488449,
0.02280823513865471,
0.05022433027625084,
0.04543105885386467,
0.034351494163274765,
0.023775150999426842,
0.0015388798201456666,
-0.02503911219537258,
-0.027300411835312843,
-0.01520982664078474,
0.015075312927365303,
-0.09263379871845245,
-0.0... |
Near-Optimal Experiment Design in Linear non-Gaussian Cyclic Models | https://openreview.net/forum?id=opAU0pYlcP | [
"Ehsan Sharifian",
"Saber Salehkaleybar",
"Negar Kiyavash"
] | Spotlight | probabilistic_methods | We study the problem of causal structure learning from a combination of observational and interventional data generated by a linear non-Gaussian structural equation model that might contain cycles. Recent results show that using mere observational data identifies the causal graph only up to a permutation-equivalence cl... | [
"Causal Discovery",
"Adaptive Experiment Design",
"Linear Non-Gaussian SCMs",
"Cyclic Causal Models",
"Adaptive Submodularity",
"Greedy Optimization"
] | null | 28,463 | 2509.21423 | title_snapshot | [
-0.01964418962597847,
-0.006088566035032272,
-0.028109274804592133,
0.024882450699806213,
0.0316486656665802,
0.017954975366592407,
0.04264211654663086,
0.01216234639286995,
-0.009552393108606339,
-0.04439178481698036,
0.0016005391953513026,
0.01570313237607479,
-0.05526602268218994,
-0.00... |
Hybrid Latent Representations for PDE Emulation | https://openreview.net/forum?id=Hh8ebJYQs3 | [
"Ali Can Bekar",
"Siddhant Agarwal",
"Christian Hüttig",
"Nicola Tosi",
"David S. Greenberg"
] | Poster | machine_learning_for_sciences | For classical PDE solvers, adjusting the spatial resolution and time step offers a trade-off between speed and accuracy. Neural emulators often achieve better speed-accuracy trade-offs by operating on a compact representation of the PDE system. Coarsened PDE fields are a simple and effective representation, but cannot ... | [
"PDE Integration",
"Physics Informed Learning",
"Neural PDE Solvers"
] | null | 28,454 | null | null | [
-0.03201768174767494,
-0.013289285823702812,
0.000849443138577044,
0.057495370507240295,
0.020964568480849266,
0.045149777084589005,
-0.003925954457372427,
-0.006356779485940933,
-0.04279346391558647,
-0.0725630670785904,
0.006168844643980265,
-0.027976248413324356,
-0.029446477070450783,
... |
MoodAngels: A Retrieval-augmented Multi-agent Framework for Psychiatry Diagnosis | https://openreview.net/forum?id=AWU93F6Bup | [
"Mengxi Xiao",
"Ben Liu",
"He Li",
"Jimin Huang",
"Qianqian Xie",
"Xiaofen Zong",
"Mang Ye",
"Min Peng"
] | Poster | applications | The application of AI in psychiatric diagnosis faces significant challenges, including the subjective nature of mental health assessments, symptom overlap across disorders, and privacy constraints limiting data availability. To address these issues, we present MoodAngels, the first specialized multi-agent framework for... | [
"psychiatry diagnosis",
"multi-agent framework",
"mental health"
] | null | 28,444 | 2506.03750 | title_snapshot | [
-0.008312089368700981,
-0.013180018402636051,
0.005827715620398521,
0.012781135737895966,
0.020540591329336166,
0.009427572600543499,
0.030561568215489388,
-0.00034578502527438104,
-0.023420561105012894,
-0.036369260400533676,
-0.018684836104512215,
0.0075701461173594,
-0.06422004848718643,
... |
Meta-learning how to Share Credit among Macro-Actions | https://openreview.net/forum?id=cJlgdpEFx9 | [
"Ionel Hosu",
"Traian Rebedea",
"Razvan Pascanu"
] | Poster | reinforcement_learning | One proposed mechanism to improve exploration in reinforcement learning is the use of macro-actions, a form of temporal abstractions over actions.
Paradoxically though, in many scenarios the naive addition of macro-actions does not lead to better exploration, but rather the opposite.
In this work, we argue that the di... | [
"deep reinforcement learning",
"macro-actions",
"exploration"
] | We propose MASP, a meta-learned similarity-based regularization for RL with macro-actions. MASP improves exploration, credit assignment, and transfer across tasks, outperforming Rainbow DQN in challenging benchmarks. | 28,441 | 2506.13690 | title_snapshot | [
-0.015161708928644657,
-0.01472823228687048,
0.0020892350003123283,
0.03033202514052391,
0.025802841410040855,
-0.005936609115451574,
0.013706624507904053,
-0.015949267894029617,
-0.03560847043991089,
-0.016895750537514687,
0.001983569236472249,
0.013592164032161236,
-0.04469430819153786,
... |
Mitigating Spurious Features in Contrastive Learning with Spectral Regularization | https://openreview.net/forum?id=TPMsCus3r0 | [
"Naghmeh Ghanooni",
"Waleed Mustafa",
"Dennis Wagner",
"Sophie Fellenz",
"Anthony Widjaja Lin",
"Marius Kloft"
] | Poster | general_machine_learning | Neural networks generally prefer simple and easy-to-learn features. When these features are spuriously correlated with the labels, the network's performance can suffer, particularly for underrepresented classes or concepts. Self-supervised representation learning methods, such as contrastive learning, are especially pr... | [
"Contrastive Learning",
"Spurious Correlation",
"Self-supervised Learning",
"Representation Learning"
] | We introduce a regularizer that promotes diverse, task-relevant features over spurious ones in contrastive learning. | 28,422 | null | null | [
-0.013962289318442345,
-0.014687794260680676,
-0.017620915547013283,
0.04634907469153404,
0.026263399049639702,
0.011284314095973969,
0.04394957050681114,
-0.017057878896594048,
-0.05128110945224762,
-0.035691406577825546,
-0.021317141130566597,
0.02400808036327362,
-0.06333177536725998,
0... |
FedFACT: A Provable Framework for Controllable Group-Fairness Calibration in Federated Learning | https://openreview.net/forum?id=6lCY5bLW8E | [
"Li Zhang",
"Zhongxuan Han",
"XiaoHua Feng",
"Jiaming Zhang",
"Yuyuan Li",
"Chaochao Chen"
] | Poster | social_and_economic_aspects_of_machine_learning | With emerging application of Federated Learning (FL) in decision-making scenarios, it is imperative to regulate model fairness to prevent disparities across sensitive groups (e.g., female, male).
Current research predominantly focuses on two concepts of group fairness within FL: *Global Fairness* (overall model dispari... | [
"Federated Learning; Fairness; Multi-Class Classification"
] | A controllable federated group-fairness calibration framework that achieves global and local fairness in multi-class classification with theoretical guarantees. | 28,415 | 2506.03777 | title_snapshot | [
0.0023427417036145926,
-0.03304971009492874,
0.015204541385173798,
0.033906545490026474,
0.029181603342294693,
0.015599466860294342,
-0.0014052625047042966,
0.0031772595830261707,
-0.027989260852336884,
-0.04763302579522133,
0.019903050735592842,
0.011056723073124886,
-0.07884660363197327,
... |
Transformer Key-Value Memories Are Nearly as Interpretable as Sparse Autoencoders | https://openreview.net/forum?id=2TKEGTfQBd | [
"Mengyu Ye",
"Jun Suzuki",
"Tatsuro Inaba",
"Tatsuki Kuribayashi"
] | Poster | social_and_economic_aspects_of_machine_learning | Recent interpretability work on large language models (LLMs) has been increasingly dominated by a feature-discovery approach with the help of proxy modules. Then, the quality of features learned by, e.g., sparse auto-encoders (SAEs), is evaluated.
This paradigm naturally raises a critical question: do such learned feat... | [
"interpretability",
"key-value memories",
"sparse autoencoders"
] | We find that transformer key-value memories are nearly as interpretable as SAE features | 28,407 | 2510.22332 | title_snapshot | [
-0.025415141135454178,
-0.027600234374403954,
-0.001977072563022375,
0.02525477297604084,
0.03765900433063507,
0.04009442776441574,
0.03649493679404259,
0.006612740457057953,
-0.030977590009570122,
-0.016138212755322456,
-0.000525444105733186,
0.01490064337849617,
-0.033005230128765106,
0.... |
Beyond Oracle: Verifier-Supervision for Instruction Hierarchy in Reasoning and Instruction-Tuned LLMs | https://openreview.net/forum?id=IQ513IX1G5 | [
"Sian-Yao Huang",
"Li-Hsien Chang",
"Che-Yu Lin",
"Cheng-Lin Yang"
] | Poster | deep_learning | Large language models (LLMs) are often prompted with multi-level directives, such as system instructions and user queries, that imply a hierarchy of authority. Yet models frequently fail to enforce this structure, especially in multi-step reasoning where errors propagate across intermediate steps. Existing methods rely... | [
"instruction hierarchy",
"verifiable supervision",
"reasoning LLMs",
"instruction-tuned LLMs",
"programmatic verification",
"oracle-free alignment",
"safety generalization"
] | We align instruction-tuned and reasoning LLMs on instruction hierarchy via executable verifier supervision, enabling oracle-free and trace-free training that generalizes to safety benchmarks. | 28,389 | null | null | [
0.008270330727100372,
-0.0025509994011372328,
-0.032303743064403534,
0.022821737453341484,
0.06093983352184296,
0.006781377829611301,
0.018457666039466858,
0.0025910083204507828,
-0.017331911250948906,
-0.008363946340978146,
-0.03485829755663872,
0.04225011542439461,
-0.0509549081325531,
0... |
ASDSV: Multimodal Generation Made Efficient with Approximate Speculative Diffusion and Speculative Verification | https://openreview.net/forum?id=IIGiVRKJYa | [
"Kaijun Zhou",
"Xingyu Yan",
"Xingda Wei",
"Xijun Li",
"Jinyu Gu"
] | Poster | infrastructure | Diffusion in transformer is central to advances in high-quality multimodal generation
but suffer from high inference latency due to their iterative nature.
Inspired by speculative decoding's success in accelerating large language models,
we propose Approximate Speculative Diffusion with Speculative Verification (ASD... | [
"Speculative Diffusion",
"Diffusion model",
"Multimodel Generation",
"Inference acceleration"
] | null | 28,386 | null | null | [
-0.018723759800195694,
-0.01542153861373663,
-0.02051769569516182,
0.0730561912059784,
0.025291655212640762,
0.05734054744243622,
0.0015478942077606916,
0.0028854107949882746,
-0.030339406803250313,
-0.031271226704120636,
-0.0003301992255728692,
0.005114284344017506,
-0.050529781728982925,
... |
Towards Accurate Time Series Forecasting via Implicit Decoding | https://openreview.net/forum?id=gqoeQPhQcE | [
"Xinyu Li",
"Yuchen Luo",
"Hao Wang",
"Haoxuan Li",
"Liuhua Peng",
"Feng Liu",
"Yandong Guo",
"Kun Zhang",
"Mingming Gong"
] | Poster | applications | Recent booming time series models have demonstrated remarkable forecasting performance. However, these methods often place greater focus on more effectively modelling the historical series, largely neglecting the forecasting phase, which generates long-term forecasts by separately predicting multiple time points. Given... | [
"Time Series Forecasting",
"Frequency-domain"
] | null | 28,364 | null | null | [
-0.011335761286318302,
-0.0328575074672699,
-0.008687308989465237,
0.007238016463816166,
0.05741429328918457,
0.05031488463282585,
0.02843903936445713,
0.01825825497508049,
-0.04579252749681473,
-0.048518624156713486,
0.0007011664565652609,
0.01603825017809868,
-0.07009127736091614,
0.0351... |
SAFEPATH: Preventing Harmful Reasoning in Chain-of-Thought via Early Alignment | https://openreview.net/forum?id=vIaNnnQxcl | [
"Wonje Jeung",
"Sangyeon Yoon",
"Minsuk Kahng",
"Albert No"
] | Poster | social_and_economic_aspects_of_machine_learning | Large Reasoning Models (LRMs) have become powerful tools for complex problem solving, but their structured reasoning pathways can lead to unsafe outputs when exposed to harmful prompts. Existing safety alignment methods reduce harmful outputs but can degrade reasoning depth, leading to significant trade-offs in complex... | [
"Large Reasoning Models (LRMs)",
"Chain-of-Thought Reasoning",
"Safety Alignment",
"Zero-shot Alignment"
] | We propose SAFEPATH, a lightweight method that aligns Large Reasoning Models to detect and suppress harmful chain-of-thought reasoning by injecting a brief safety signal at the start of reasoning. | 28,363 | 2505.14667 | title_snapshot | [
-0.007771593984216452,
-0.026569204404950142,
-0.02223081886768341,
0.048921242356300354,
0.057061731815338135,
-0.005828120745718479,
0.0394938588142395,
-0.015442629344761372,
-0.021475164219737053,
-0.020661432296037674,
-0.011386962607502937,
0.023010235279798508,
-0.06659120321273804,
... |
Diffusing DeBias: Synthetic Bias Amplification for Model Debiasing | https://openreview.net/forum?id=66Z5tS8E45 | [
"Massimiliano Ciranni",
"Vito Paolo Pastore",
"Roberto Di Via",
"Enzo Tartaglione",
"Francesca Odone",
"Vittorio Murino"
] | Poster | applications | The effectiveness of deep learning models in classification tasks is often challenged by the quality and quantity of training data whenever they are affected by strong spurious correlations between specific attributes and target labels. This results in a form of bias affecting training data, which typically leads to un... | [
"Model debiasing",
"bias amplification",
"diffusion models",
"image classification"
] | null | 28,359 | 2502.09564 | title_snapshot | [
0.03226722776889801,
-0.03423743695020676,
-0.02703118324279785,
0.04896391183137894,
0.032535944133996964,
0.03073902800679207,
0.020578932017087936,
0.005660282913595438,
-0.010350818745791912,
-0.0735756978392601,
0.042224153876304626,
-0.00992879644036293,
-0.08428999036550522,
0.01693... |
Escaping saddle points without Lipschitz smoothness: the power of nonlinear preconditioning | https://openreview.net/forum?id=7qrhHzZpTA | [
"Alexander Bodard",
"Panagiotis Patrinos"
] | Spotlight | optimization | We study generalized smoothness in nonconvex optimization, focusing on $(L_0, L_1)$-smoothness and anisotropic smoothness. The former was empirically derived from practical neural network training examples, while the latter arises naturally in the analysis of nonlinearly preconditioned gradient methods. We introduce a ... | [
"Nonconvex optimization",
"generalized smoothness",
"saddle point avoidance"
] | null | 28,346 | 2509.15817 | title_snapshot | [
-0.07426927238702774,
-0.04665801301598549,
0.042599622160196304,
0.055748507380485535,
0.03614921495318413,
0.02100336365401745,
0.035524044185876846,
-0.005721653811633587,
-0.03527195006608963,
-0.07035741955041885,
-0.009727311320602894,
-0.016408074647188187,
-0.049162525683641434,
-0... |
Perturb a Model, Not an Image: Towards Robust Privacy Protection via Anti-Personalized Diffusion Models | https://openreview.net/forum?id=5XoqKCmkS7 | [
"Tae-Young Lee",
"Juwon Seo",
"Jong Hwan Ko",
"Gyeong-Moon Park"
] | Poster | social_and_economic_aspects_of_machine_learning | Recent advances in diffusion models have enabled high-quality synthesis of specific subjects, such as identities or objects. This capability, while unlocking new possibilities in content creation, also introduces significant privacy risks, as personalization techniques can be misused by malicious users to generate unau... | [
"text-to-image diffusion models",
"personalization",
"privacy"
] | We prevent unauthorized personalization of diffusion models at the model level. | 28,338 | 2511.01307 | title_snapshot | [
0.021115649491548538,
-0.025520792230963707,
0.022212445735931396,
0.0469900481402874,
0.05751285329461098,
0.01508302055299282,
0.03020387887954712,
-0.0331413634121418,
-0.00446266820654273,
-0.057837046682834625,
-0.01051950454711914,
-0.02619972825050354,
-0.04516493156552315,
0.006052... |
Why Playing Against Diverse and Challenging Opponents Speeds Up Coevolution: A Theoretical Analysis on Combinatorial Games | https://openreview.net/forum?id=wWSVjaVZBu | [
"Alistair Benford",
"Per Kristian Lehre"
] | Poster | theory | Competitive coevolutionary algorithms (CoEAs) have a natural application to problems that are adversarial or feature strategic interaction. However, there is currently limited theoretical insight into how to avoid pathological behaviour associated with CoEAs. In this paper we use impartial combinatorial games as a chal... | [
"coevolution",
"runtime analysis",
"combinatorial games"
] | null | 28,329 | null | null | [
-0.010457312688231468,
-0.03333473950624466,
-0.021217545494437218,
0.012797764502465725,
0.03576672822237015,
0.03643140569329262,
0.05081917718052864,
-0.0010643063578754663,
-0.015096706338226795,
-0.04075590521097183,
0.01823369413614273,
0.008201820775866508,
-0.080351822078228,
-0.00... |
Theoretical Guarantees for the Retention of Strict Nash Equilibria by Coevolutionary Algorithms | https://openreview.net/forum?id=e5QEGDVsqn | [
"Alistair Benford",
"Per Kristian Lehre"
] | Poster | theory | Most methods for finding a Nash equilibrium rely on procedures that operate over the entire action space, making them infeasible for settings with too many actions to be searched exhaustively. Randomised search heuristics such as coevolutionary algorithms offer benefits in such settings, however they lack many of the t... | [
"coevolution",
"adversarial optimisation",
"Nash equilibria"
] | null | 28,310 | null | null | [
-0.03563179820775986,
-0.011434774845838547,
-0.03205705061554909,
0.03500022739171982,
0.02472066693007946,
0.03490469977259636,
0.030862702056765556,
0.010900956578552723,
-0.01212792843580246,
-0.04451708123087883,
0.018020251765847206,
0.014766702428460121,
-0.07752686738967896,
-0.033... |
Stochastic Principal-Agent Problems: Computing and Learning Optimal History-Dependent Policies | https://openreview.net/forum?id=u0rNHqMpFD | [
"Jiarui Gan",
"R Majumdar",
"Debmalya Mandal",
"Goran Radanovic"
] | Poster | theory | We study a stochastic principal-agent model. A principal and an agent interact in a stochastic environment, each privy to observations about the state not available to the other. The principal has the power of commitment, both to elicit information from the agent and to signal her own information. The players communica... | [
"stochastic games",
"Markov games",
"Stackelberg games",
"information design",
"mechanism design"
] | null | 28,300 | null | null | [
-0.05844838544726372,
-0.02539471536874771,
-0.004351119045168161,
0.037844933569431305,
0.023755226284265518,
0.026773449033498764,
0.012490510009229183,
0.006126685068011284,
-0.02646685019135475,
-0.05333114415407181,
-0.01390637457370758,
-0.007647774647921324,
-0.06252267956733704,
-0... |
Large language models can learn and generalize steganographic chain-of-thought under process supervision | https://openreview.net/forum?id=2g5cJqX15Y | [
"Robert McCarthy",
"Joey SKAF",
"Luis Ibanez-Lissen",
"Vasil Georgiev",
"Connor Watts",
"Hannes Whittingham",
"Lorena Gonzalez-Manzano",
"Cameron Tice",
"Edward James Young",
"Puria Radmard",
"David Lindner"
] | Poster | deep_learning | Chain-of-thought (CoT) reasoning not only enhances large language model performance but also provides critical insights into decision-making processes, marking it as a useful tool for monitoring model intent and planning. By proactively preventing models from acting on CoT indicating misaligned or harmful intent, CoT m... | [
"AI Safety",
"AI Control",
"Steganography",
"Encoded Reasoning",
"Chain-of-Thought",
"Reinforcement Learning",
"LLMs"
] | We show that penalizing certain CoT reasoning makes LLMs learn encoding schemes that generalize to unseen examples. | 28,295 | 2506.01926 | title_snapshot | [
0.0007843549828976393,
-0.025133149698376656,
-0.03301617503166199,
0.060358114540576935,
0.05374053120613098,
-0.011713428422808647,
0.02929839678108692,
0.022503361105918884,
-0.026044512167572975,
-0.019115300849080086,
-0.04364557936787605,
0.028348034247756004,
-0.06609127670526505,
-... |
PyraMotion: Attentional Pyramid-Structured Motion Integration for Co-Speech 3D Gesture Synthesis | https://openreview.net/forum?id=QJSrgYcf4b | [
"Zhizhuo Yin",
"Yuk Hang Tsui",
"Pan Hui"
] | Poster | applications | Generating full-body human gestures encompassing face, body, hands, and global movements from audio is crucial yet challenging for virtual avatar creation. Existing systems tokenize gestures frame-wise, predicting tokens of each frame from the input audio. However, expressive human gestures consist of varied patterns w... | [
"Co-Speech Motion Synthesis",
"3D",
"Generation",
"Representation Learning"
] | null | 28,272 | null | null | [
0.00344588584266603,
0.01125289872288704,
0.002990658860653639,
0.010031165555119514,
-0.010274984873831272,
0.06609185039997101,
0.05782222002744675,
0.020241696387529373,
-0.03285994380712509,
-0.045299723744392395,
-0.034631337970495224,
-0.0267224982380867,
-0.0550735667347908,
-0.0115... |
Monoculture or Multiplicity: Which Is It? | https://openreview.net/forum?id=DO5LtJc80w | [
"Mila Gorecki",
"Moritz Hardt"
] | Poster | social_and_economic_aspects_of_machine_learning | Two narratives about machine learning ecosystems grew out of recent algorithmic fairness discourse. In one, dubbed \emph{monoculture}, algorithmic ecosystems tend toward homogeneity akin to a single model making all decisions. Individuals then face the risk of systematic exclusion with no recourse. In the other, \emph{... | [
"monoculture",
"multiplicity",
"large language models"
] | We systematically evaluate the concerns of multiplicity and monoculture in a suite of large language models and prediction tasks. | 28,266 | null | null | [
-0.01979857124388218,
-0.02822471596300602,
-0.04678323492407799,
0.029976464807987213,
0.01949462853372097,
0.03471710905432701,
0.026744071394205093,
0.029137471690773964,
-0.03821345418691635,
-0.009378916583955288,
-0.012438826262950897,
0.01047731563448906,
-0.08286634087562561,
-0.00... |
Scalable Evaluation and Neural Models for Compositional Generalization | https://openreview.net/forum?id=heQsyrMDzm | [
"Giacomo Camposampiero",
"Pietro Barbiero",
"Michael Hersche",
"Roger Wattenhofer",
"Abbas Rahimi"
] | Poster | evaluation | Compositional generalization—a key open challenge in modern machine learning—requires models to predict unknown combinations of known concepts. However, assessing compositional generalization remains a fundamental challenge due to the lack of standardized evaluation protocols and the limitations of current benchmarks, ... | [
"compositional generalization",
"compositionality",
"disentanglement",
"representation learning",
"computer vision"
] | We introduce a novel, scalable framework to evaluate compositional generalization, leverage it to evaluate more than 5k models, and propose a family of neural models pushing the Pareto frontier on this task. | 28,258 | 2511.02667 | title_snapshot | [
0.0046073030680418015,
-0.004571173340082169,
0.003173581790179014,
0.03895588964223862,
0.018487922847270966,
0.03241710737347603,
0.00870794989168644,
0.02193848229944706,
-0.01753961853682995,
-0.02070850506424904,
-0.025477958843111992,
0.001454512239433825,
-0.07349075376987457,
0.009... |
MuRating: A High Quality Data Selecting Approach to Multilingual Large Language Model Pretraining | https://openreview.net/forum?id=jHWCeU39Ft | [
"Zhixun Chen",
"Ping Guo",
"Wenhan Han",
"Yifan Zhang",
"BINBINLIU",
"Haobin Lin",
"Fengze Liu",
"Yan Zhao",
"Bingni Zhang",
"Taifeng Wang",
"Yin Zheng",
"Trevor Cohn",
"Meng Fang"
] | Poster | deep_learning | Data quality is a critical driver of large language model performance, yet existing model-based selection methods focus almost exclusively on English, neglecting other languages that are essential in the training mix for multilingual LLMs. We introduce MuRating, a scalable framework that transfers high-quality English ... | [
"Natural Language Processing"
] | We proposed a new data selection method for pretraining multilingual Large Language Models | 28,254 | 2507.01785 | title_snapshot | [
-0.03127444535493851,
-0.03642828390002251,
-0.012033581733703613,
0.050446346402168274,
0.016313429921865463,
0.023302851244807243,
0.0009924726327881217,
0.014308450743556023,
-0.021690065041184425,
-0.014608797617256641,
-0.02043004147708416,
0.047497496008872986,
-0.05108259990811348,
... |
Diffusion Adaptive Text Embedding for Text-to-Image Diffusion Models | https://openreview.net/forum?id=cHi8QxGrZH | [
"Byeonghu Na",
"Minsang Park",
"Gyuwon Sim",
"Donghyeok Shin",
"HeeSun Bae",
"Mina Kang",
"Se Jung Kwon",
"Wanmo Kang",
"Il-chul Moon"
] | Poster | deep_learning | Text-to-image diffusion models rely on text embeddings from a pre-trained text encoder, but these embeddings remain fixed across all diffusion timesteps, limiting their adaptability to the generative process. We propose Diffusion Adaptive Text Embedding (DATE), which dynamically updates text embeddings at each diffusio... | [
"Diffusion models",
"text-to-image diffusion models",
"text-to-image generation"
] | We propose Diffusion Adaptive Text Embedding (DATE), which improves text-to-image diffusion models by dynamically refining text embeddings throughout the diffusion sampling process. | 28,249 | 2510.23974 | title_snapshot | [
-0.0024332778993993998,
-0.029526205733418465,
0.004703255835920572,
0.06536367535591125,
0.06130869314074516,
0.03395982086658478,
0.037032052874565125,
0.00790075957775116,
0.0063545117154717445,
-0.04306380823254585,
-0.008501576259732246,
-0.0061968485824763775,
-0.03414333611726761,
0... |
Activation Control for Efficiently Eliciting Long Chain-of-thought Ability of Language Models | https://openreview.net/forum?id=XNo4yS9n1k | [
"Zekai Zhao",
"Qi Liu",
"Kun Zhou",
"Zihan Liu",
"Yifei Shao",
"Zhiting Hu",
"Biwei Huang"
] | Spotlight | deep_learning | Despite the remarkable reasoning performance, eliciting the long chain-of-thought(CoT) ability in large language models(LLMs) typically requires costly reinforcement learning or supervised fine-tuning on high-quality distilled data. We investigate the internal mechanisms behind this capability and show that a small set... | [
"Large Language Models",
"Long Chain of Thoughts"
] | null | 28,241 | 2505.17697 | title_snapshot | [
-0.025866936892271042,
-0.030396610498428345,
-0.018210362643003464,
0.03910122439265251,
0.0325789637863636,
0.01432811189442873,
0.01830550841987133,
0.005412273574620485,
-0.012870745733380318,
0.00687514990568161,
-0.029015392065048218,
0.042654119431972504,
-0.058277592062950134,
-0.0... |
LinEAS: End-to-end Learning of Activation Steering with a Distributional Loss | https://openreview.net/forum?id=EBONa3tT3K | [
"Pau Rodriguez",
"Michal Klein",
"Eleonora Gualdoni",
"Valentino Maiorca",
"Arno Blaas",
"Luca Zappella",
"marco cuturi",
"Xavier Suau"
] | Poster | deep_learning | The growing use of generative models in daily life calls for efficient mechanisms to control their generation, to e.g. produce safe content or provide users with tools to explore style changes. Ideally, such mechanisms should require low volume of unpaired data (\ie without explicit preference), and should be cheap, bo... | [
"controllability",
"generative models",
"toxicity",
"images"
] | We propose an inference-time intervention framework based on Optimal Transport that generalizes previous methods and allows interpretable control of both LLMs and Diffusion models. | 28,240 | 2503.10679 | title_snapshot | [
0.0031220291275531054,
-0.01368432305753231,
-0.028273215517401695,
0.011706323362886906,
0.025316037237644196,
0.016159843653440475,
0.03555016592144966,
0.007292478810995817,
-0.007572580128908157,
-0.003888308070600033,
-0.03308653458952904,
0.021425645798444748,
-0.05159293860197067,
-... |
PDEfuncta: Spectrally-Aware Neural Representation for PDE Solution Modeling | https://openreview.net/forum?id=NfBrMDF0Xi | [
"Minju Jo",
"Woojin Cho",
"Uvini Balasuriya Mudiyanselage",
"Seungjun Lee",
"Noseong Park",
"Kookjin Lee"
] | Poster | machine_learning_for_sciences | Scientific machine learning often involves representing complex solution fields that exhibit high-frequency features such as sharp transitions, fine-scale oscillations, and localized structures. While implicit neural representations (INRs) have shown promise for continuous function modeling, capturing such high-frequen... | [
"SciML",
"meta-learning",
"data compression"
] | null | 28,235 | 2506.12790 | title_snapshot | [
-0.03435872122645378,
-0.03395257890224457,
0.009270130656659603,
0.03187175095081329,
0.01620279625058174,
0.02509007602930069,
0.024927761405706406,
0.0006305829738266766,
-0.04070058465003967,
-0.0375363752245903,
0.014503560960292816,
0.005129874683916569,
-0.06954696774482727,
0.01757... |
Minimizing False-Positive Attributions in Explanations of Non-Linear Models | https://openreview.net/forum?id=ORrCEtiiVX | [
"Anders Gjølbye",
"Stefan Haufe",
"Lars Kai Hansen"
] | Poster | social_and_economic_aspects_of_machine_learning | Suppressor variables can influence model predictions without being dependent on the target outcome, and they pose a significant challenge for Explainable AI (XAI) methods. These variables may cause false-positive feature attributions, undermining the utility of explanations. Although effective remedies exist for linear... | [
"Explainable AI",
"Interpretability",
"Suppressor Variables",
"Non-Linear Problems",
"Machine Learning",
"EEG"
] | PatternLocal is a novel XAI method that refines local linearization approaches to reduce false-positive feature attributions in non-linear explanations. | 28,232 | 2505.11210 | title_snapshot | [
-0.010220223106443882,
0.016696332022547722,
-0.0070641557686030865,
0.03391473367810249,
0.0245608352124691,
0.034916702657938004,
0.02170848660171032,
-0.02128124050796032,
-0.03310719132423401,
-0.05198537930846214,
-0.014257116243243217,
0.017371002584695816,
-0.06101401150226593,
0.00... |
Rethinking PCA Through Duality | https://openreview.net/forum?id=IFQBrEAuQ6 | [
"Jan Quan",
"Johan Suykens",
"Panagiotis Patrinos"
] | Poster | general_machine_learning | Motivated by the recently shown connection between self-attention and (kernel) principal component analysis (PCA), we revisit the fundamentals of PCA. Using the difference-of-convex (DC) framework, we present several novel formulations and provide new theoretical insights. In particular, we show the kernelizability and... | [
"PCA",
"DCA",
"Non-Convex Optimization",
"Unsupervised Learning"
] | null | 28,229 | 2510.18130 | title_snapshot | [
-0.02744041010737419,
-0.022658616304397583,
0.012575606815516949,
0.03979954123497009,
0.032981593161821365,
0.043306492269039154,
0.0028030695393681526,
0.01415929477661848,
-0.015860198065638542,
-0.04724512994289398,
-0.03899513557553291,
-0.008935126475989819,
-0.05385934188961983,
-0... |
A Markov Decision Process for Variable Selection in Branch & Bound | https://openreview.net/forum?id=05Svr0k5C9 | [
"Paul STRANG",
"Zacharie ALES",
"Côme Bissuel",
"Olivier Juan",
"Safia Kedad-Sidhoum",
"Emmanuel Rachelson"
] | Poster | optimization | Mixed-Integer Linear Programming (MILP) is a powerful framework used to address a wide range of NP-hard combinatorial optimization problems, often solved by Branch and bound (B&B). A key factor influencing the performance of B&B solvers is the variable selection heuristic governing branching decisions. Recent contribut... | [
"Mixed-integer linear programming",
"Branch and bound",
"Reinforcement learning",
"Markov decision process"
] | We enhance the potential for reinforcement learning applications in mixed-integer linear programming by modeling variable selection in Branch and Bound as a Markov decision process. | 28,221 | 2510.19348 | title_snapshot | [
-0.038119200617074966,
0.014159563928842545,
-0.02563927322626114,
0.04017361253499985,
0.05790845677256584,
0.04281461611390114,
0.02239576168358326,
-0.024145497009158134,
-0.04738501086831093,
-0.022570470348000526,
-0.008880614303052425,
-0.00248670089058578,
-0.09128915518522263,
-0.0... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.