ICML
Collection
Accepted papers for ICML (International Conference on Machine Learning), one dataset per year. • 14 items • Updated
paper_id stringlengths 10 10 | title stringlengths 18 172 | paper_url stringlengths 42 42 | authors listlengths 1 59 | type stringclasses 2
values | primary_area stringclasses 84
values | abstract large_stringlengths 288 2.49k | keywords listlengths 1 25 | TL;DR large_stringlengths 11 250 ⌀ | submission_number int64 1 34.8k | arxiv_id stringlengths 10 10 ⌀ | arxiv_id_source stringclasses 2
values |
|---|---|---|---|---|---|---|---|---|---|---|---|
oiMjaUbSWp | On the Epistemic Uncertainty of Overparametrized Neural Networks | https://openreview.net/forum?id=oiMjaUbSWp | [
"David Rügamer"
] | Poster | probabilistic_methods | Epistemic uncertainty is often viewed as a reducible uncertainty that vanishes with increasing data. This perspective implicitly assumes parameter identifiability and equates epistemic uncertainty with predictive variability. In overparametrized neural networks, however, model parameters are typically non-identifiable ... | [
"Epistemic Uncertainty",
"Overparametrized Models",
"Identifiability"
] | null | 1 | 2605.25234 | title_snapshot |
vSzRJyg6k0 | Reward-free Alignment for Conflicting Objectives | https://openreview.net/forum?id=vSzRJyg6k0 | [
"Peter Chen",
"Xiaopeng Li",
"Xi Chen",
"Tianyi Lin"
] | Spotlight | deep_learning->large_language_models | Direct alignment methods are increasingly used to align large language models (LLMs) with human preferences. However, many real-world alignment problems involve multiple conflicting objectives, where naive aggregation of preferences can lead to unstable training and poor trade-offs. In particular, weighted loss methods... | [
"Multi-Objective Alignment",
"Multi-Objective Optimization",
"LLM Preference Alignment",
"AI Safety"
] | null | 2 | 2602.02495 | title_snapshot |
Jva4wVEySO | Envisioning Beyond the Few: Disentangled Semantics and Primitives for Few-Shot Atypical Layout-to-Image Generation | https://openreview.net/forum?id=Jva4wVEySO | [
"Nan Bao",
"Yifan Zhao",
"Wenzhuang Wang",
"Jia Li"
] | Poster | deep_learning->generative_models_and_autoencoders | The layout-to-image (L2I) task enables fine-grained control over image generation via object categories and spatial layouts. However, existing L2I methods yield fragmented and distorted generations under few-shot atypical settings. We term this failure as representation fragmentation, arising from a granularity mismatc... | [
"layout-to-image generation",
"few-shot image generation",
"diffusion models",
"atypical visual domains"
] | A few-shot layout-to-image generation framework for atypical visual domains. | 4 | 2605.31266 | title_snapshot |
WUK8JIeetF | One-step Latent-free Image Generation with Pixel Mean Flows | https://openreview.net/forum?id=WUK8JIeetF | [
"Yiyang Lu",
"Susie Lu",
"Qiao Sun",
"Hanhong Zhao",
"Zhicheng Jiang",
"Xianbang Wang",
"Tianhong Li",
"Zhengyang Geng",
"Kaiming He"
] | Poster | deep_learning->generative_models_and_autoencoders | Modern diffusion/flow-based models for image generation typically exhibit two core characteristics: (i) using multi-step sampling, and (ii) operating in a latent space. Recent advances have made encouraging progress on each aspect individually, paving the way toward one-step diffusion/flow without latents. In this work... | [
"Diffusion Models",
"Flow Matching"
] | one-step latent-free image generation | 6 | 2601.22158 | title_snapshot |
5TiuerrwR8 | Olaf-World: Orienting Latent Actions for Video World Modeling | https://openreview.net/forum?id=5TiuerrwR8 | [
"Yuxin Jiang",
"Yuchao Gu",
"Ivor Tsang",
"Mike Zheng Shou"
] | Poster | applications | Scaling action-controllable world models is limited by the scarcity of action labels. While latent action learning promises to extract control interfaces from unlabeled video, learned latents often fail to transfer across contexts: they entangle scene-specific cues and lack a shared coordinate system. This occurs becau... | [
"World Model",
"Latent Action",
"Video Generation"
] | null | 11 | 2602.10104 | title_snapshot |
53wE3EbrgK | REALISTA: Realistic Latent Adversarial Attacks that Elicit LLM Hallucinations | https://openreview.net/forum?id=53wE3EbrgK | [
"Buyun Liang",
"Jinqi Luo",
"Liangzu Peng",
"Kwan Ho Ryan Chan",
"Darshan Thaker",
"Kaleab A Kinfu",
"Fengrui Tian",
"Hamed Hassani",
"Rene Vidal"
] | Poster | social_aspects->safety | Large language models (LLMs) achieve strong performance across many tasks but remain vulnerable to hallucinations, making it important to systematically evaluate their reliability under realistic adversarial inputs. We formulate hallucination elicitation as a constrained optimization problem, where the goal is to find ... | [
"LLM hallucination",
"safety",
"trustworthiness"
] | We propose REALISTA, a realistic latent adversarial attack framework that elicits LLM hallucinations by optimizing continuous combinations of semantically equivalent and coherent editing directions. | 15 | 2605.12813 | title_snapshot |
O8jabXEYlQ | Phase-Aware Mixture of Experts for Agentic Reinforcement Learning | https://openreview.net/forum?id=O8jabXEYlQ | [
"Shengtian Yang",
"Yu Li",
"Shuo He",
"Yewen Li",
"Qingpeng Cai",
"Peng Jiang",
"Lei Feng"
] | Poster | reinforcement_learning->everything_else | Reinforcement learning (RL) has equipped LLM agents with a strong ability to solve complex tasks. However, existing RL methods normally use a single policy network, causing simplicity bias where simple tasks occupy most parameters and dominate gradient updates, leaving insufficient capacity for complex tasks.
A plausib... | [
"Agentic RL",
"MoE",
"Phase-Aware"
] | Agentic RL, MoE, Phase-Aware | 18 | 2602.17038 | title_snapshot |
CT2tSmahVQ | Deep sequence models tend to memorize geometrically; it is unclear why | https://openreview.net/forum?id=CT2tSmahVQ | [
"Shahriar Noroozizadeh",
"Vaishnavh Nagarajan",
"Elan Rosenfeld",
"Sanjiv Kumar"
] | Poster | deep_learning->large_language_models | Deep sequence models are said to store atomic facts predominantly in the form of *associative* memory: a brute-force lookup of co-occurring entities. We identify a dramatically different form of storage of atomic facts that we term as *geometric* memory. Here, the model has synthesized embeddings encoding novel *global... | [
"Geometric Memory",
"Next-Token Prediction",
"Associative Memory",
"Memorization",
"In-Context",
"In-Weights"
] | Deep sequence models can solve an in-weights path-finding task where they failed in-context by implicitly organizing local facts into a coherent geometric structure, challenging the theory that they store knowledge as simple associations. | 22 | 2510.26745 | title_snapshot |
JcjRShiRQz | Dissect and Prune: Enhancing Robustness in AI-Generated Image Detection | https://openreview.net/forum?id=JcjRShiRQz | [
"Dahye Kim",
"Jaehyun Choi",
"Hyun Seok Seong",
"Seongho Kim",
"Donghun Lee",
"Sungwon Yi",
"Jang-Ho Choi"
] | Poster | social_aspects->safety | While existing AI-generated image detectors report high performance, we identify that this is largely driven by a critical *prediction asymmetry*: a bias toward the real class that severely limits sensitivity to generated content, especially under standard post-processing operations such as compression and resizing. We... | [
"AI-generated Image Detection",
"Synthetic Image Detection",
"Image Forensics"
] | We propose DEAR, a method that addresses the prediction asymmetry in AI-generated image detection by pruning spurious features via inpainted images, significantly enhancing robustness against unseen models and perturbations. | 27 | 2606.10309 | title_snapshot |
3NQSeJOfkz | SVRG and Beyond via Posterior Correction | https://openreview.net/forum?id=3NQSeJOfkz | [
"Nico Daheim",
"Thomas Möllenhoff",
"Ming Liang Ang",
"Mohammad Emtiyaz Khan"
] | Spotlight | probabilistic_methods->variational_inference | Stochastic Variance Reduced Gradient (SVRG) and its variants aim to speed-up training by using gradient corrections. Originally proposed over a decade ago, these methods have never been connected to any Bayesian method at a fundamental level. Here, we fill this gap and derive surprising new connections of SVRG to a rec... | [
"variational learning",
"bayesian deep learning",
"variance reduction",
"stochastic optimization",
"convex optimization"
] | We generalize SVRG using Bayes and derive new methods. | 39 | 2512.01930 | title_snapshot |
vMcu1h3fOV | Joint Model and Data Sparsification via the Marginal Likelihood | https://openreview.net/forum?id=vMcu1h3fOV | [
"Alexander Timans",
"Thomas Möllenhoff",
"Christian A. Naesseth",
"Mohammad Emtiyaz Khan",
"Eric Nalisnick"
] | Poster | probabilistic_methods | Sparse recovery in linear systems underpins applications from signal processing to high-dimensional regression. Sparse Bayesian Learning, grounded in the principle of automatic relevance determination (ARD), offers a practical Bayesian mechanism for feature sparsity via marginal likelihood optimization. Yet, its relian... | [
"sparsity",
"marginal likelihood",
"sparse bayesian learning"
] | We learn both per-feature and per-sample relevance scores via a single marginal likelihood objective. | 42 | 2605.29908 | title_snapshot |
kTOxGQJkwt | TABX: A High-Throughput Sandbox Battle Simulator for Multi-Agent Reinforcement Learning | https://openreview.net/forum?id=kTOxGQJkwt | [
"Hayeong Lee",
"JunHyeok Oh",
"Byung-Jun Lee"
] | Poster | reinforcement_learning->multiagent | The design of environments plays a critical role in shaping the development and evaluation of cooperative multi-agent reinforcement learning (MARL) algorithms. While existing benchmarks highlight critical challenges, they often lack the modularity required to design custom evaluation scenarios. We introduce the Totally... | [
"Reinforcement Learning",
"Multi Agent Reinforcement Learning",
"JAX"
] | null | 45 | 2602.01665 | title_snapshot |
D2evvc90tF | Can Microcanonical Langevin Dynamics Leverage Mini-Batch Gradient Noise? | https://openreview.net/forum?id=D2evvc90tF | [
"Emanuel Sommer",
"Kangning Diao",
"Jakob Robnik",
"Uros Seljak",
"David Rügamer"
] | Poster | probabilistic_methods->monte_carlo_and_sampling_methods | Scaling inference methods such as Markov chain Monte Carlo to high-dimensional models remains a central challenge in Bayesian deep learning. A promising recent proposal, microcanonical Langevin Monte Carlo, has shown state-of-the-art performance across a wide range of problems. However, its reliance on full-dataset gra... | [
"Microcanonical Langevin",
"Sampling",
"Bayesian Deep Learning"
] | null | 46 | 2602.06500 | title_snapshot |
aPnQpmwHW7 | OmniSIFT: Modality-Asymmetric Token Compression for Efficient Omni-modal Large Language Models | https://openreview.net/forum?id=aPnQpmwHW7 | [
"Yue Ding",
"Yiyan Ji",
"Jungang Li",
"Xuyang Liu",
"Xinlong Chen",
"Junfei Wu",
"Bozhou Li",
"Bohan Zeng",
"Yang Shi",
"Yushuo Guan",
"Yuanxing Zhang",
"Jiaheng Liu",
"Qiang Liu",
"Pengfei Wan",
"Liang Wang"
] | Poster | deep_learning->large_language_models | Omni-modal Large Language Models (Omni-LLMs) have demonstrated strong capabilities in audio-video understanding tasks. However, their reliance on long multimodal token sequences leads to substantial computational overhead. Despite this challenge, token compression methods designed for Omni-LLMs remain limited. To bridg... | [
"efficiency",
"multimodal",
"omni"
] | We propose a new compression framework for omni-modal large language models. | 49 | 2602.04804 | title_snapshot |
EhJ1R2N6sn | Geometric Coherence Learning for Structuring Value Functions in Plain MDPs | https://openreview.net/forum?id=EhJ1R2N6sn | [
"Zuyuan Zhang",
"Zeyu Fang",
"Tian Lan"
] | Poster | reinforcement_learning | Geometric properties can be leveraged to stabilize and speed reinforcement learning. Existing examples include encoding symmetry structure, geometry-aware data augmentation, and enforcing structural restrictions. In this paper, we take a novel view of RL through the lens of order theory and recast value function estima... | [
"Symmetry",
"logic-order",
"Geometric"
] | null | 58 | 2602.02978 | title_judge |
FFAHL32Wok | Decomposing the Basic Abilities of Large Language Models: Mitigating Cross-Task Interference in Multi-Task Instruct-Tuning | https://openreview.net/forum?id=FFAHL32Wok | [
"Bing Wang",
"Ximing Li",
"Changchun Li",
"Jinjin Chi",
"Gang Niu",
"Masashi Sugiyama"
] | Poster | general_machine_learning->transfer_multitask_and_metalearning | Recently, the prominent performance of large language models (LLMs) has been largely driven by multi-task instruct-tuning. Unfortunately, this training paradigm suffers from a key issue, named cross-task interference, due to conflicting gradients over shared parameters among different tasks. Some previous methods mitig... | [
"large language model",
"instruct-tuning",
"multi-task learning",
"mixture-of-expert"
] | null | 61 | 2605.05676 | title_snapshot |
3qq7rvJzoj | HodgeFlow Policy Search by Topologically Dissecting Temporal-Difference Signals in Non-Markovian Environments | https://openreview.net/forum?id=3qq7rvJzoj | [
"Zuyuan Zhang",
"Sizhe Tang",
"Tian Lan"
] | Poster | reinforcement_learning | Non-Markovian dynamics are commonly found in real-world environments due to long-range dependencies, partial observability, and memory effects. The Bellman equation that is the central pillar of Reinforcement learning (RL) becomes only approximately valid under Non-Markovian. Existing work often focus on practical algo... | [
"HodgeFlow",
"topological projection"
] | null | 75 | null | null |
ZRymA5ujIZ | AffIn-Space: Learning Affine-Invariant Representations for 3D Spatial Understanding with MLLMs | https://openreview.net/forum?id=ZRymA5ujIZ | [
"Zhenyu Lu",
"Liupeng Li",
"Jinpeng Wang",
"Haoqian Kang",
"Manyuan Zhang",
"Yan Feng",
"Ke Chen",
"Yaowei Wang"
] | Poster | deep_learning->large_language_models | While MLLMs show promising capacity on general visual understanding, they suffer from *geometric fragility*: standard visual representations often degrade rapidly under changes in viewpoint and viewing distance.
Our analysis identifies that existing paradigms, whether relying on input-level fusion or latent reconstruct... | [
"multimodal fusion",
"affine invariant representations",
"spatial reasoning",
"spatial grounding"
] | AffIn-Space addresses the geometric fragility in existing 3D MLLMs by enforcing strict affine invariance through explicit geometric resampling and orthogonal projection constraints , achieving state-of-the-art results on benchmarks. | 81 | null | null |
hEQDZg1ZeD | Are First-Order Diffusion Samplers Really Slower? A Fast Forward-Value Approach | https://openreview.net/forum?id=hEQDZg1ZeD | [
"Yuchen Jiao",
"Na Li",
"Changxiao Cai",
"Gen Li"
] | Poster | deep_learning->generative_models_and_autoencoders | Higher-order ODE solvers have become a standard tool for accelerating diffusion probabilistic model (DPM) sampling, motivating the widespread view that first-order methods are inherently slower and that increasing discretization order is the primary path to faster generation. This paper challenges this belief and revis... | [
"Diffusion models",
"image generation",
"DDIM"
] | null | 82 | 2512.24927 | title_snapshot |
TxHOT32Maj | Social Hippocampus Memory Learning | https://openreview.net/forum?id=TxHOT32Maj | [
"Liping Yi",
"Zhiming Zhao",
"Kewen Zhu",
"Xiang Li",
"Zhiwei Shang",
"Qinghua Hu"
] | Poster | social_aspects | Social learning highlights that learning agents improve not in isolation, but through interaction and structured knowledge exchange with others. When introduced into machine learning, this principle gives rise to social machine learning (SML), where multiple agents collaboratively learn by sharing abstracted knowledge.... | [
"Social machine learning",
"hippocampus memory learning",
"short--long memory",
"individual--collective memory"
] | null | 83 | 2603.25614 | title_snapshot |
8WA5l5tjF6 | CVSearch: Empowering Multimodal LLMs with Cognitive Visual Search for High-Resolution Image Perception | https://openreview.net/forum?id=8WA5l5tjF6 | [
"Liupeng Li",
"Haoqian Kang",
"Zhenyu Lu",
"Jinpeng Wang",
"Bin Chen",
"Ke Chen",
"Yaowei Wang"
] | Poster | deep_learning->large_language_models | High-resolution (HR) image perception presents a key bottleneck for multimodal large language models (MLLMs).
While visual search offers a promising solution, existing methods struggle with the trade-off between coverage and efficiency. Visual expert-assisted search is efficient but prone to blind spots when proposals ... | [
"Multimodal Large Language Models",
"High-Resolution Image Perception",
"Cognitive Visual Search",
"Semantic-Aware Scanning"
] | CVSearch solves the MLLM high-resolution bottleneck via a training-free cognitive visual search framework that optimizes semantic-aware scanning using Semantic Guided Adaptive Patching and Dynamic Bottom-Up Search. | 87 | 2605.23655 | title_snapshot |
wjQw9XVSEt | Making Learner Weakness Actionable for Learning from Demonstration with Novice Teachers | https://openreview.net/forum?id=wjQw9XVSEt | [
"Yuqing Zhu",
"Matthew Howard"
] | Poster | applications->robotics | Learning from demonstration can be an effective way to teach robots task-oriented policies. However, in an interactive setting when demonstrations are limited by time or other budgetary constraints, it is challenging to find those that fix the learner's (remaining) errors. This is especially difficult for novice teache... | [
"Learning from Demonstration",
"Human-Robot Interaction",
"Demonstration Efficiency",
"Novice Teachers",
"Robotics"
] | null | 93 | null | null |
lA85cqhMh0 | Revisiting Uncertainty: On Evidential Learning for Partially Relevant Video Retrieval | https://openreview.net/forum?id=lA85cqhMh0 | [
"Jun Li",
"Peifeng Lai",
"Xuhang Lou",
"Jinpeng Wang",
"Yuting Wang",
"Ke Chen",
"Yaowei Wang",
"Shu-Tao Xia"
] | Poster | applications->computer_vision | Partially relevant video retrieval aims to retrieve untrimmed videos using text queries that describe only partial content. However, the inherent asymmetry between brief queries and rich video content inevitably introduces uncertainty into the retrieval process.
In this setting, vague queries often induce semantic am... | [
"partially relevant video retrieval",
"video-text retrieval",
"cross-modal retrieval",
"uncertainty estimation",
"evidential deep learning"
] | We propose Holmes, a hierarchical evidential learning framework for PRVR that models query ambiguity via fine-grained inter-video query category identification and flexible optimal transport–based intra-video temporal supervision. | 98 | 2605.06083 | title_snapshot |
xpIUvw8ANa | Learning, Solving and Optimizing PDEs with TensorGalerkin: an efficient high-performance Galerkin assembly algorithm | https://openreview.net/forum?id=xpIUvw8ANa | [
"Shizheng Wen",
"Mingyuan chi",
"Tianwei Yu",
"Ben Moseley",
"Mike Yan Michelis",
"Pu Ren",
"Hao Sun",
"Siddhartha Mishra"
] | Poster | applications->chemistry_physics_and_earth_sciences | We present a unified algorithmic framework for the numerical solution, constrained optimization, and physics-informed learning of PDEs with a variational structure. Our framework is based on a Galerkin discretization of the underlying variational forms, and its high efficiency stems from a novel highly-optimized and GP... | [
"AI for Science",
"Physics-Informed Machine Learning",
"Operator Learning",
"Partial Differential Equations",
"Differentiable Simulation",
"PDE Constrained Optimization",
"Finite Element Methods"
] | TensorGalerkin, a unified PyTorch framework that reformulates Galerkin assembly as a strictly tensorized Map-Reduce operation, enabling highly efficient, end-to-end differentiable PDE solving, operator learning, and inverse design on GPUs. | 106 | 2602.05052 | title_snapshot |
3W2IAC4cNh | Beyond Reactivity: Proactive Adaptive Conformal Inference for Online LLM Factuality | https://openreview.net/forum?id=3W2IAC4cNh | [
"Xinyu Liu",
"Jun Wu"
] | Poster | deep_learning->large_language_models | Large Language Models (LLMs) often produce hallucinated outputs, which limit
their reliability in high-stakes applications. Conformal prediction can
provide guarantees on the correctness and factuality of LLM outputs, but existing
approaches rely on the exchangeability assumption, which rarely holds in
online settings ... | [
"Conformal Prediction",
"LLM Hallucination",
"Uncertainty Quantification",
"Distribution Shift"
] | PACE addresses the challenge of maintaining valid LLM factuality under online distribution shifts by dynamically adjusting the conformal prediction step size | 109 | null | null |
bkcxLMm3Gp | DenseSteer: Steering Small Language Models towards Dense Math Reasoning | https://openreview.net/forum?id=bkcxLMm3Gp | [
"Yang Ouyang",
"Shuhang Lin",
"Jung-Eun Kim"
] | Poster | deep_learning->large_language_models | Large language models (LLMs) demonstrate strong chain-of-thought (CoT) reasoning abilities, while smaller models ($\leq$ 3B parameters) significantly underperform on multi-step reasoning tasks. Based on empirical analyses of the Qwen-2.5 model family on math reasoning benchmarks, we find that more proficient reasoning ... | [
"Math Reasoning",
"CoT",
"LLM",
"Activation Steering"
] | DenseSteer improves small language models by steering them to produce more compact reasoning steps without retraining. | 111 | 2605.29247 | title_snapshot |
t5BgTJ7Z0k | Self-Refining Video Sampling | https://openreview.net/forum?id=t5BgTJ7Z0k | [
"Sangwon Jang",
"Taekyung Ki",
"Jaehyeong Jo",
"Saining Xie",
"Jaehong Yoon",
"Sung Ju Hwang"
] | Poster | deep_learning->generative_models_and_autoencoders | Modern video generators still struggle with complex physical dynamics, often falling short of physical realism. Existing approaches address this using external verifiers or additional training on augmented data, which is computationally expensive and still limited in capturing fine-grained motion. In this work, we pres... | [
"Video generation",
"Self-refinement",
"Flow matching",
"Physical realism"
] | a training-free video sampling method for better physics and motion dynamics | 114 | 2601.18577 | title_snapshot |
2NRrsz4nUB | From Abstraction to Instantiation: Learning Behavioral Representation for Vision-Language-Action Model | https://openreview.net/forum?id=2NRrsz4nUB | [
"Bing Hu",
"Zaijing Li",
"Rui Shao",
"Junda Chen",
"April Hua Liu",
"Wei-Shi Zheng",
"Liqiang Nie"
] | Spotlight | applications->robotics | Vision-Language-Action (VLA) models often suffer from performance degradation under distribution shifts, as they struggle to learn generalized behavior representations across varying environments. While existing approaches attempt to construct behavior representations through action-centric latent variables, they are o... | [
"Robotic Manipulation",
"Vision-Language-Action Model",
"Flow Matching"
] | null | 118 | 2605.22671 | title_snapshot |
xeH37plPS3 | Order Matters in Retrosynthesis: Structure-aware Generation via Reaction-Center-Guided Discrete Flow Matching | https://openreview.net/forum?id=xeH37plPS3 | [
"Chenguang Wang",
"Zihan Zhou",
"LEI BAI",
"Tianshu Yu"
] | Poster | applications->chemistry_physics_and_earth_sciences | Template-free retrosynthesis methods treat the task as black-box sequence generation, limiting learning efficiency, while semi-template approaches rely on rigid reaction libraries that constrain generalization. We address this gap with a key insight: atom ordering in neural representations matters.
Building on this in... | [
"retrosynthesis",
"discrete flow matching",
"generative models"
] | We show that encoding reaction centers as positional inductive bias outperforms brute-force scaling: a 280K-parameter model with proper ordering matches a 65M model, achieving SOTA with 6× faster training and 10× fewer sampling steps. | 124 | 2602.13136 | title_snapshot |
4v5iCXWmcR | Resting Neurons, Active Insights: Robustifying Activation Sparsity in LLMs via Spontaneity | https://openreview.net/forum?id=4v5iCXWmcR | [
"Haotian Xu",
"Jiannan Yang",
"Tian Gao",
"Tsui-Wei Weng",
"Tengfei Ma"
] | Poster | social_aspects->everything_else | Activation sparsity offers a compelling route to accelerate large language model (LLM) inference by selectively suppressing hidden activations, yet existing approaches exhibit severe accuracy degradation at high sparsity. We show that this failure stems from representational instability: *activation sparsity disrupts i... | [
"Activation Sparsity",
"Efficient AI",
"LLM Architecture Design"
] | null | 127 | 2512.12744 | title_snapshot |
jgebUtw1lA | Turning Drift into Constraint: Robust Reasoning Alignment in Non-Stationary Multi-Stream Environments | https://openreview.net/forum?id=jgebUtw1lA | [
"Xiaoyu Yang",
"En Yu",
"Wei Duan",
"Jie Lu"
] | Poster | deep_learning->theory | This paper identifies a critical yet underexplored challenge in reasoning alignment from multiple multi-modal large language models (MLLMs): In non-stationary environments, the diverse reasoning distributions of source models often evolve unpredictably, transmitting systematic biases and drift to the target model. To a... | [
"Concept Drift",
"Multi Modal Large Language Model",
"Robust Medical Diagnosis"
] | null | 129 | 2510.04142 | title_snapshot |
UgPfadSsaA | Distributionally Robust Markov Games with Average Reward | https://openreview.net/forum?id=UgPfadSsaA | [
"Zachary Andrew Roch",
"Yue Wang"
] | Poster | reinforcement_learning->multiagent | We propose and study distributionally robust Markov games (DR‑MGs) with the average‑reward criterion as a crucial framework for multi-agent decision-making under model mismatches and over extended horizons. Under a standard irreducible assumption, we first derive a correspondence between the optimal policies and the so... | [
"Markov games",
"distributional robustness",
"average reward"
] | This paper establishes the existence of stationary Nash Equilibria in average-reward distributionally robust Markov games and introduces two provably convergent algorithms to guarantee multi-agent resilience against environmental uncertainty. | 130 | 2508.03136 | title_snapshot |
LShQYeeop5 | GIFT: Bootstrapping Image-to-CAD Program Synthesis via Geometric Feedback | https://openreview.net/forum?id=LShQYeeop5 | [
"Giorgio Giannone",
"Anna C. Doris",
"Amin Heyrani Nobari",
"Kai Xu",
"Akash Srivastava",
"Faez Ahmed"
] | Poster | applications->everything_else | Generating executable CAD programs from images requires alignment between visual geometry and symbolic program representations, a capability that current methods fail to learn reliably as design complexity increases. Existing fine-tuning approaches rely on either limited supervised datasets or expensive post-training p... | [
"inference-time-scaling",
"cad-program-synthesis",
"geometric-inference-feedback-tuning",
"compute-driven-data-augmentation"
] | Amortizing Inference-Time Data Augmentation for CAD Program Synthesis | 131 | 2603.27448 | title_snapshot |
Y0Dhwhu8Uc | PASA: A Principled Embedding-Space Watermarking Approach for LLM-Generated Text under Semantic-Invariant Attacks | https://openreview.net/forum?id=Y0Dhwhu8Uc | [
"Zhenxin Ai",
"Haiyun He"
] | Poster | deep_learning->large_language_models | Watermarking for large language models (LLMs) is a promising approach for detecting LLM-generated text and enabling responsible deployment. However, existing watermarking methods are often vulnerable to semantic-invariant attacks, such as paraphrasing.
We propose PASA, a principled, robust, and distortion-free watermar... | [
"LLM Watermarking; Theoretically Grounded; Robust; Distortion-Free; Semantic-Invariant Attacks"
] | A principled, robust and distortion-free watermarking approach on latent semantic space. | 137 | 2605.10977 | title_snapshot |
DfZS0M8leJ | Generalized Linear Bandits with Memory | https://openreview.net/forum?id=DfZS0M8leJ | [
"Heesang Ann",
"Hyunjun Choi",
"Taehyun Hwang",
"Younghoon Shin",
"Haeju Cheong",
"Min-hwan Oh"
] | Poster | theory->online_learning_and_bandits | We study generalized linear bandits with memory, an endogenous non-stationary setting in which rewards depend on past actions through a finite memory matrix. Building on prior work for linear models Clerici et al.,(2024), we show that the previously known $\tilde{\mathcal{O}}(T^{3/4})$ regret bound stems from a loose a... | [
"Generalized linear model",
"bandit with memory",
"non-stationary"
] | We introduce the generalized linear bandit with memory and propose a novel algorithm with provable guarantees for this setting. | 138 | null | null |
qjwFbN77kx | FiSeR: Fine-Grained Source Representations for Cross-Domain AI Image Detection | https://openreview.net/forum?id=qjwFbN77kx | [
"Shan Zhang",
"Yongxin He",
"Mingming Zhang",
"Huiwen Tian",
"Lei Ma"
] | Poster | applications->computer_vision | Real-world synthetic image detectors often generalize poorly under domain shift despite strong in-domain performance. Using unsupervised UMAP projections, we find that natural and synthetic features remain partially separable on unseen datasets, yet performance still drops, suggesting that the classification head overf... | [
"AI-generated image detection; representation learning; cross-domain generalization; contrastive learning;"
] | null | 144 | 2606.00606 | title_snapshot |
cCMbxJcDmH | PhoStream: Benchmarking Real-World Streaming for Omnimodal Assistants in Mobile Scenarios | https://openreview.net/forum?id=cCMbxJcDmH | [
"Xudong Lu",
"Guan Huankang",
"Yang Bo",
"Jinpeng Chen",
"Xintong Guo",
"Shuhan LI",
"Fang Liu",
"Peiwen Sun",
"Xueying Li",
"Wei Zhang",
"Xue Yang",
"Rui Liu",
"Hongsheng Li"
] | Poster | applications->computer_vision | Multimodal Large Language Models excel at offline audio-visual understanding, but their ability to serve as mobile assistants in continuous real-world streams remains underexplored. In daily phone use, mobile assistants must track streaming audio-visual inputs and respond at the right time, yet existing benchmarks are ... | [
"Benchmark",
"Streaming Video Understanding",
"Multimodal Large Language Models",
"Mobile Assistant"
] | We introduce PhoStream, a mobile-centric streaming benchmark that unifies on-screen and off-screen scenarios for evaluating video, audio, and temporal reasoning. | 159 | 2601.22575 | title_snapshot |
iwS6Lg1Uwx | Is Spurious Correlation Removal Always Learnable? | https://openreview.net/forum?id=iwS6Lg1Uwx | [
"Yibo Zhou",
"Bo Li",
"Hai-Miao Hu",
"Hanzi Wang",
"Xiaokang Zhang",
"Ruifan Zhang"
] | Poster | theory->everything_else | Invariant learning can fail even when the invariant structure is statistically identifiable. We show a conditional computational barrier: under a black-box samplable supervised sparse recovery primitive motivated by average-case sparse-recovery reductions, there exist \emph{samplable} multi-environment instances with a... | [
"spurious correlation removal",
"computational-statistical gap"
] | Prove (under Planted Clique) invariant feature recovery can be statistically identifiable yet polynomial-time hard, and show environment diversity controls identifiability and sample complexity. | 162 | 2606.12930 | title_snapshot |
GAQE4Wr53f | Row-Stochastic Matrices Can Provably Outperform Doubly Stochastic Matrices in Decentralized Learning | https://openreview.net/forum?id=GAQE4Wr53f | [
"Bing Liu",
"Boao Kong",
"Limin Lu",
"Kun Yuan",
"Chengcheng Zhao"
] | Poster | optimization->large_scale_parallel_and_distributed | Decentralized learning often involves a weighted global loss with heterogeneous node weights $\lambda$. We revisit two natural strategies for incorporating these weights: (i) embedding them into the local losses to retain a uniform weight (and thus a doubly stochastic matrix), and (ii) keeping the original losses while... | [
"decentralized learning",
"stochastic optimization",
"row-stochastic matrix"
] | null | 166 | 2511.19513 | title_snapshot |
bQk1qnXcqM | Log-Normal Multiplicative Dynamics for Stable Low-Precision Deep Learning | https://openreview.net/forum?id=bQk1qnXcqM | [
"Keigo Nishida",
"Eren Mehmet KIRAL",
"Kenichi Bannai",
"Mohammad Emtiyaz Khan",
"Thomas Möllenhoff"
] | Poster | deep_learning->algorithms | We propose Log-Normal Multiplicative Dynamics (LMD), a new algorithm to stabilize low-precision deep learning. LMD is motivated by the robustness of biological neural networks where synaptic spine sizes follow a log-normal distribution and fluctuate via noisy multiplicative dynamics. So far, no algorithm has successful... | [
"Bayesian Deep Learning",
"Low Precision Deep Learning",
"Multiplicative Weight Update"
] | null | 167 | 2506.17768 | title_judge |
goVeG3ui1Y | Error Amplification Limits ANN-to-SNN Conversion in Continuous Control | https://openreview.net/forum?id=goVeG3ui1Y | [
"Zijie Xu",
"Zihan Huang",
"Yiting Dong",
"Kang Chen",
"Wenxuan Liu",
"Zhaofei Yu"
] | Poster | applications->neuroscience_cognitive_science | Spiking Neural Networks (SNNs) can achieve competitive performance by converting already existing well-trained Artificial Neural Networks (ANNs), avoiding further costly training. This property is particularly attractive in Reinforcement Learning (RL), where training through environment interaction is expensive and pot... | [
"Spiking Neural Networks",
"ANN-to-SNN Conversion",
"Brain-inspired Computing",
"Reinforcement Learning"
] | We identify that temporally correlated errors cause severe state drift in ANN-to-SNN conversion for continuous control, and propose a gradient-free Cross-Step Residual Potential Initialization mechanism to suppress this error amplification. | 175 | 2601.21778 | title_snapshot |
F13K6NNeAz | FRISM: Fine-Grained Reasoning Injection via Subspace-Level Model Merging for Vision–Language Models | https://openreview.net/forum?id=F13K6NNeAz | [
"Chenyu Huang",
"Peng Ye",
"Xudong Tan",
"Jinhan Mu",
"Shenghe Zheng",
"Li Shen",
"Tao Chen"
] | Poster | general_machine_learning->transfer_multitask_and_metalearning | Efficiently enhancing the reasoning capabilities of Vision-Language Models (VLMs) by merging them with Large Reasoning Models (LRMs) has emerged as a promising direction. However, existing methods typically operate at a coarse-grained layer level, which often leads to a trade-off between injecting reasoning capabilitie... | [
"model merging",
"multimodal reasoning"
] | An efficient and effective method to improve the reasoning performance of VLMs through model merging in the SVD level. | 176 | 2601.21187 | title_snapshot |
91gJ3Zp2lT | Mixture Prototype Flow Matching for Open-Set Supervised Anomaly Detection | https://openreview.net/forum?id=91gJ3Zp2lT | [
"Fuyun Wang",
"Yuanzhi Wang",
"Xu Guo",
"Sujia Huang",
"Tong Zhang",
"Dan Wang",
"Hui Yan",
"Xin Liu",
"Zhen Cui"
] | Poster | applications->computer_vision | Open-set supervised anomaly detection (OSAD) aims to identify unseen anomalies using limited anomalous supervision. However, existing prototype-based methods typically model normal data via a unimodal Gaussian prior, failing to capture inherent multi-modality and resulting in blurred decision boundaries. To address thi... | [
"Anomaly Detection",
"Diffusion Models"
] | null | 190 | 2605.02438 | title_snapshot |
26ELwHEbQL | Anomaly-Preference Image Generation | https://openreview.net/forum?id=26ELwHEbQL | [
"Fuyun Wang",
"Yuanzhi Wang",
"Xu Guo",
"Sujia Huang",
"Tong Zhang",
"Dan Wang",
"Xin Liu",
"Hui Yan",
"Zhen Cui"
] | Poster | applications->computer_vision | Synthesizing realistic and diverse anomalous samples from limited data is vital for robust model generalization. However, existing methods struggle to reconcile fidelity and diversity, often hampered by distribution misalignment and overfitting, respectively. To mitigate this, we introduce Anomaly Preference Optim... | [
"Anomaly Detection",
"Diffusion Models"
] | null | 194 | 2605.02439 | title_snapshot |
k7XzObg9Hy | FOAM: Blocked State Folding for Memory-Efficient LLM Training | https://openreview.net/forum?id=k7XzObg9Hy | [
"Ziqing Wen",
"Jiahuan Wang",
"Ping Luo",
"Dongsheng Li",
"Tao Sun"
] | Poster | deep_learning->large_language_models | Large language models (LLMs) have demonstrated remarkable performance due to their large parameter counts and extensive training data. However, their scale leads to significant memory bottlenecks during training, especially when using memory-intensive optimizers like Adam. Existing memory-efficient approaches often rel... | [
"Large language models",
"machine learning syetem",
"memory efficient training"
] | null | 200 | 2512.07112 | title_snapshot |
ZxMHQCFYr1 | Expected Returns and Policy Inconsistency-Aware Offline Federated Deep Reinforcement Learning | https://openreview.net/forum?id=ZxMHQCFYr1 | [
"Meng XU",
"Zhongying Chen",
"Weiwei Fu",
"Yan Li",
"Shuguang Wang",
"Jianping Wang"
] | Poster | reinforcement_learning->deep_rl | Offline Federated Deep Reinforcement Learning (FDRL) methods aggregate multiple client-side offline Deep Reinforcement Learning (DRL) models, each trained locally, to facilitate knowledge sharing while preserving privacy. Existing offline FDRL methods assign client weights during global aggregation using either simple ... | [
"Federated Deep Reinforcement Learning; Offline Deep Reinforcement Learning"
] | This paper proposes an offline federated deep reinforcement learning framework that evaluates the capabilities of client models and the global model by combining policy inconsistency and expected return. | 204 | null | null |
TVAMni6Cq1 | PartCo: Part-Level Correspondence Priors Enhance Category Discovery | https://openreview.net/forum?id=TVAMni6Cq1 | [
"Fernando Julio Cendra",
"Kai Han"
] | Poster | general_machine_learning->unsupervised_and_semisupervised_learning | Generalized Category Discovery (GCD) aims to identify both known and novel categories within unlabeled data by leveraging a set of labeled examples from known categories. Existing GCD methods primarily depend on semantic labels and global image representations, often overlooking the detailed part-level cues that are cr... | [
"Generalized Category Discovery"
] | We present PartCo, a framework that uses part-level visual cues to enhance Generalized Category Discovery, outperforming most existing methods by bridging the gap between semantic labels and part-level visual compositions. | 216 | 2509.22769 | title_snapshot |
WfxJlvbHpD | GeoEvo: Identity-Aware Potential Game with Geometric Evolution for Personalized Multimodal Federated Learning | https://openreview.net/forum?id=WfxJlvbHpD | [
"Chen Wang",
"Yongli Hu",
"Huajie Jiang",
"Kan Guo",
"Tengfei Liu",
"Junbin Gao",
"Yanfeng Sun",
"Baocai Yin"
] | Poster | general_machine_learning | We reconceptualize Personalized Multimodal Federated Learning (PMFL) by treating missing modalities as intrinsic structural identities that constrain each client to a distinct Riemannian submanifold, rather than as deficiencies to be compensated. To reconcile the tension between identity preservation and cross-client c... | [
"Personalized Multimodal Federated Learning; Potential Game; Riemannian Optimization"
] | null | 226 | null | null |
Du8BMHfNzp | PATRA: Pattern-Aware Alignment and Balanced Reasoning for Time Series Question Answering | https://openreview.net/forum?id=Du8BMHfNzp | [
"Junkai Lu",
"Peng Chen",
"Xingjian Wu",
"Yang Shu",
"Chenjuan Guo",
"Christian S. Jensen",
"Bin Yang"
] | Poster | applications->time_series | Time series reasoning demands both the perception of complex dynamics and logical depth. However, existing LLM-based approaches exhibit two limitations: they often treat time series merely as text or images, failing to capture the patterns like trends and seasonalities needed to answer specific questions; and when trai... | [
"LLM",
"Time Series",
"Reasoning"
] | null | 233 | 2602.23161 | title_snapshot |
JKMZyCghr4 | Adaptive Probe-based Steering for Robust LLM Jailbreaking | https://openreview.net/forum?id=JKMZyCghr4 | [
"Junxi Chen",
"Junhao Dong",
"Xiaohua Xie"
] | Poster | social_aspects->security | Recent work has demonstrated the potential of contrastive steering for jailbreaking Large Language Models (LLMs). However, existing methods rely on limited and inherently biased contrastive prompts and require laborious manual tuning of steering strength, limiting their robustness and effectiveness. In this paper, we l... | [
"Adversarial Machine Learning",
"LLM",
"Contrastive Steering",
"Jailbreaking"
] | We leverage the idea of model extraction and contrastive activations' statistics to enhance the robustness, usability, and effectiveness of probe-based contrastive steering for LLM jailbreaking. | 235 | 2605.20286 | title_snapshot |
NNGIniylxB | Efficient Multi-round LLM Inference over Disaggregated Serving | https://openreview.net/forum?id=NNGIniylxB | [
"Wenhao He",
"Youhe Jiang",
"Penghao Zhao",
"Quanqing Xu",
"Eiko Yoneki",
"Bin CUI",
"Fangcheng Fu"
] | Poster | deep_learning->large_language_models | With the rapid evolution of Large Language Models (LLMs), multi-round workflows, such as autonomous agents and iterative retrieval, have become increasingly prevalent. However, this raises hurdles for serving LLMs under prefill-decode (PD) disaggregation, a widely adopted paradigm that separates the compute-bound prefi... | [
"LLM Serving",
"Multi-round Inference",
"PD Disaggregation",
"Agentic Workflows"
] | null | 242 | 2602.14516 | title_snapshot |
5pQiAaDN7E | Return-Critic: Bridging Goal Discrepancy for Efficient Visual Reinforcement Learning | https://openreview.net/forum?id=5pQiAaDN7E | [
"Ruyi Lu",
"Xuesong Wang",
"Hengrui Zhang",
"Yuhu Cheng"
] | Poster | reinforcement_learning->deep_rl | Sample inefficiency remains a challenge in pixel-based visual reinforcement learning (RL), primarily due to ineffective state representation learning. While recent advances employ auxiliary tasks to improve representation learning, their representation goals (e.g., mask reconstruction, state prediction) are misaligned ... | [
"visual reinforcement learning; representation learning; sample efficiency"
] | Return-Critic bridges the discrepancy between representation goal and RL ultimate goal by return prediction, provably improving representation quality and achieving 68% performance boost on challenging tasks from DMControl. | 245 | null | null |
LMBt26pQj1 | COLLIE: Guiding Skill Discovery in Semantically Coherent Latent Space | https://openreview.net/forum?id=LMBt26pQj1 | [
"Yao Luan",
"Ni Mu",
"Hanfei Ge",
"Yiqin Yang",
"Bo XU",
"Qing-Shan Jia"
] | Poster | reinforcement_learning | Unsupervised skill discovery (USD) aims to learn diverse behaviors without reward functions, but often results in task-irrelevant or hazardous behaviors due to uniform exploration. Guided skill discovery (GSD) addresses this issue by incorporating human intent to focus exploration on meaningful regions. However, existi... | [
"Reinforcement Learning",
"Skill Discovery"
] | null | 258 | 2606.00950 | title_snapshot |
xGjhMGmmCF | Taylor-Gaussians-Flow: Towards Non-uniform Motion for Novel View Synthesis from Monocular Video | https://openreview.net/forum?id=xGjhMGmmCF | [
"Zaoming Yan",
"Qizhou Chen",
"Yaomin Huang",
"Pengcheng Lei",
"Chenhao Shi",
"Yi Xu",
"Haichuan Song",
"Faming Fang"
] | Poster | applications->computer_vision | Long-term non-uniform motion poses a significant challenge for Novel View Synthesis (\textbf{NVS}), as it requires modeling higher-order motion, such as acceleration.
Existing methods primarily rely on deformation fields or scene flow, which are limited to first-order approximations.
Due to neglecting higher-order mot... | [
"Video Frame Generation",
"Gaussian Splatting."
] | null | 261 | null | null |
tIG4D44jn9 | Discrete Diffusion with Physical Mass Constraints for \emph{De Novo} Peptide Sequencing | https://openreview.net/forum?id=tIG4D44jn9 | [
"An Zeyu",
"Wanyu Lin"
] | Poster | applications->health_medicine | $\textit{De novo}$ peptide sequencing is a pivotal technique that directly reconstructs amino acid sequences from tandem mass spectrometry (MS/MS) data; it enables the identification of novel proteins and variants absent from reference databases. Previous methods are typically based on autoregressive (AR) decoding or o... | [
"De Novo Peptide Sequencing",
"Discrete Diffusion",
"Precursor Mass",
"Spectral Fragment Ion",
"Global Reasoning and Iterative Refinement"
] | We introduce PhysNovo, a generative paradigm that shifts peptide sequencing from translation to physically constrained iterative reasoning via discrete diffusion, ensuring strict validity and superior generalization. | 268 | null | null |
UB6M4BSvwJ | DynVLA: Learning World Dynamics for Action Reasoning in Autonomous Driving | https://openreview.net/forum?id=UB6M4BSvwJ | [
"Shuyao Shang",
"Bing Zhan",
"Yunfei Yan",
"Yuqi Wang",
"Yingyan Li",
"Yasong An",
"Xiaoman Wang",
"Jierui Liu",
"Lu Hou",
"Lue Fan",
"Zhaoxiang Zhang",
"Tieniu Tan"
] | Poster | applications->robotics | We propose DynVLA, a driving VLA model that introduces a new CoT paradigm termed Dynamics CoT. DynVLA forecasts compact world dynamics before action generation, enabling more informed and physically grounded decision-making. To obtain compact dynamics representations, DynVLA introduces a Dynamics Tokenizer that compres... | [
"Autonomous Driving",
"End-to-End Autonomous Driving",
"Vision–Language–Action Model",
"Chain-of-Thought"
] | DynVLA enables VLA models to reason over future world dynamics, improving decision quality while preserving inference efficiency. | 269 | 2603.11041 | title_snapshot |
f4k0XsOlCZ | Rel-MOSS: Towards Imbalanced Relational Deep Learning on Relational Databases | https://openreview.net/forum?id=f4k0XsOlCZ | [
"Jun Yin",
"Peng Huo",
"Bangguo Zhu",
"Hao Yan",
"Senzhang Wang",
"Shirui Pan",
"Chengqi Zhang"
] | Poster | deep_learning->graph_neural_networks | In recent advances, to enable a fully data-driven learning paradigm on relational databases (RDB), relational deep learning (RDL) is proposed to structure the RDB as a heterogeneous entity graph and adopt the graph neural network (GNN) as the predictive model. However, existing RDL methods neglect the imbalance problem... | [
"Relational Deep Learning",
"Class Imbalance"
] | Rel-MOSS is a novel graph neural network framework that solves the severe class imbalance problem in relational databases by intelligently filtering out majority noise and synthesizing structurally faithful minority examples. | 270 | 2603.07916 | title_snapshot |
doy2uIaAYt | Tracing the Persona Circuit: How Large Language Models Encode and Express Character Traits | https://openreview.net/forum?id=doy2uIaAYt | [
"Guanzheng Qin",
"Chenghao Sun",
"Zhining Xie",
"Xinmei Tian"
] | Poster | social_aspects->accountability_transparency_and_interpretability | Large Language Models (LLMs) demonstrate remarkable potential in role-playing tasks but frequently suffer from personality decay—termed "Out-of-Character" (OOC) behavior—during prolonged interactions.
While heuristic strategies exist to align model behaviors, the internal computational dynamics driving personality exp... | [
"Large Language Models",
"activation patching"
] | null | 275 | null | null |
7UEBX1KU1y | Conditional Equivalence of DPO and RLHF: Assumptions, Failure Modes, and Provable Alignment | https://openreview.net/forum?id=7UEBX1KU1y | [
"Zhiqin Yang",
"Yonggang Zhang",
"Wei Xue",
"Dong Fang",
"Bo Han",
"Yike Guo"
] | Spotlight | applications->language_speech_and_dialog | Direct Preference Optimization (DPO) has emerged as a popular alternative to Reinforcement Learning from Human Feedback (RLHF), offering theoretical equivalence with a simpler implementation. We prove this equivalence is _conditional_ rather than universal, depending on an implicit assumption frequently violated in pra... | [
"Preference Optimization",
"Reinforcement Learning from Human Feedback"
] | null | 276 | 2605.20834 | title_judge |
3tN36gT4tB | MoSSP: A Momentum-Based Single-Loop Stochastic Penalty Method for Nonconvex Constrained DC-regularized Optimization | https://openreview.net/forum?id=3tN36gT4tB | [
"Luxuan Li",
"Chunfeng Cui",
"Xiao Wang"
] | Poster | optimization->nonconvex | In this paper, we study a structured class of nonconvex constrained stochastic problems with difference-of-convex (DC) regularization, where the feasible set is possibly nonconvex and the concave part of the DC regularizer is allowed to be nonsmooth.
The fundamental challenge lies in maintaining feasibility for nonconv... | [
"Nonsmooth optimization",
"DC optimization",
"Constrained optimization",
"Penalty methods",
"Smoothing approximation",
"Momentum",
"Oracle complexity"
] | This paper develops an efficient single-loop stochastic penalty method for nonconvex constrained DC-regularized optimization with provable complexity guarantee. | 279 | 2605.29635 | title_snapshot |
kVBNDXHVeo | Mode Seeking meets Mean Seeking for Fast Long Video Generation | https://openreview.net/forum?id=kVBNDXHVeo | [
"Shengqu Cai",
"Weili Nie",
"Chao Liu",
"Julius Berner",
"Lvmin Zhang",
"Nanye Ma",
"Hansheng Chen",
"Maneesh Agrawala",
"Leonidas Guibas",
"Gordon Wetzstein",
"Arash Vahdat"
] | Poster | deep_learning->generative_models_and_autoencoders | Scaling video generation from seconds to minutes faces a critical bottleneck: while short-video data is abundant and high-fidelity, coherent long-form data is scarce and limited to narrow domains.
While multi-resolution image training works because higher resolution is largely an interpolation of the same underlying pa... | [
"Video Generation",
"Representation Learning"
] | null | 280 | 2602.24289 | title_snapshot |
LnNbhU8IcR | FlowPET: Physics-Informed Symplectic Flow Matching for Low-Count PET Reconstruction | https://openreview.net/forum?id=LnNbhU8IcR | [
"Zheng Zhang",
"Hao Tang",
"Yingying Hu",
"zhanli hu",
"Jing Qin"
] | Poster | applications->health_medicine | Low-count Positron Emission Tomography (PET) reconstruction is severely hindered by the dissipative nature of prevailing generative models, where the inherent phase-space contraction leads to the numerical extinction (``wash-out'') of weak but diagnostically critical lesion signals. To overcome this geometric limitatio... | [
"Medical Imaging",
"Positron Emission Tomography",
"Low-Count PET Reconstruction"
] | null | 282 | null | null |
EW7FmahpLs | Motion Dynamics Learning for Few-Shot Embodied Adaptation | https://openreview.net/forum?id=EW7FmahpLs | [
"Sibo He",
"Weiying Xie",
"Daixun Li",
"Junhao Zhong",
"Jiayun Tian",
"Yunke Wang",
"Leyuan Fang",
"Gang He",
"Yunsong Li"
] | Poster | applications->computer_vision | Vision-Language-Action (VLA) models have shown strong potential for robotic manipulation, yet adapting pretrained models to novel tasks typically relies on substantial task-specific demonstrations, limiting scalability.
Current VLA methods mostly focus on action imitation, which ignores the richer structure contained... | [
"System Application"
] | null | 287 | null | null |
Ai3a79cTvr | SVL: Empowering Spiking Neural Networks for Efficient 3D Open-World Understanding | https://openreview.net/forum?id=Ai3a79cTvr | [
"Xuerui Qiu",
"Shaowei Gu",
"Peixi Wu",
"JiaKui Hu",
"Yaozhi Wen",
"Yuqi Pan",
"Xinhao Luo",
"Bo XU",
"Guoqi Li"
] | Spotlight | applications->neuroscience_cognitive_science | Spiking Neural Networks (SNNs) offer an energy--efficient route to 3D spatio--temporal perception, yet they lag behind Artificial Neural Networks (ANNs) due to weak pretraining and heavy inference stacks, limiting generalization and multimodal reasoning (e.g., zero--shot 3D classification and open--world QA). We presen... | [
"Vision-language models; Spike-driven; Spike Point Transformer;Spiking Neural Network;"
] | The first spike-based multimodal framework that empowers SNNs with open-world 3D perception while maintaining spike-driven efficiency. | 290 | 2505.17674 | title_judge |
bniyv9QWYc | RTPrune: Reading-Twice Inspired Token Pruning for Efficient DeepSeek-OCR Inference | https://openreview.net/forum?id=bniyv9QWYc | [
"Ben Wan",
"Yan Feng",
"Zihan Tang",
"Weizhe Huang",
"Yuting Zeng",
"Jia Wang",
"Tongxuan Liu"
] | Poster | applications->computer_vision | DeepSeek-OCR leverages visual–text compression to reduce long-text processing costs and accelerate inference, yet visual tokens remain prone to redundant textual and structural information. Moreover, current token pruning methods for conventional vision–language models (VLMs) fail to preserve textual fidelity due to im... | [
"DeepSeek-OCR",
"Visual Token Pruning",
"Inference Acceleration",
"Visual–Text Compression"
] | RTPrune is a two-stage, training-free visual token pruning framework for DeepSeek-OCR that mimics the LLM’s reading-twice behavior and adopts a dynamic pruning ratio, achieving efficient inference while preserving high OCR accuracy. | 291 | 2605.00392 | title_snapshot |
mMlt3tQ7Ot | AdamO: A Collapse-Suppressed Optimizer for Offline RL | https://openreview.net/forum?id=mMlt3tQ7Ot | [
"Nan Qiao",
"Sheng Yue",
"Shuning Wang",
"Ju Ren"
] | Poster | reinforcement_learning->batchoffline | Offline reinforcement learning (RL) can fail spectacularly when bootstrapped temporal-difference (TD) updates amplify their own errors, driving the critic toward extreme and unusable Q-values.
A key counterintuitive insight of this work is that collapse is not only a property of the backup rule or network architecture... | [
"Offline RL",
"Reinforcement Learning;",
"Value Function Collapse"
] | Supressing value collapse in temporal-difference updates through a decoupled orthogonality correction regulated by a strict task-alignment budget. | 300 | 2605.01968 | title_snapshot |
khMSeuLLhh | Adversarial Attack and Defense for Denoising Diffusion Sampling | https://openreview.net/forum?id=khMSeuLLhh | [
"Zhao-Rong Lai",
"Xiwen Yuan",
"Jian Weng"
] | Poster | social_aspects->security | Denoising diffusion sampling (DDS) is an emerging approach for generating new samples that have the same distribution as some training samples. However, it is vulnerable to adversarial attacks by even a Gaussian perturbation. In this work, we propose a complete set of adversarial attack and defense methodology for DDS.... | [
"Denoising diffusion sampling",
"adversarial attack and defense",
"local variation"
] | We propose an adversarial attack and defense approach for denoising diffusion sampling. | 303 | null | null |
Ga3AR4EF6R | Milestone-Guided Policy Learning for Long-Horizon Language Agents | https://openreview.net/forum?id=Ga3AR4EF6R | [
"Zixuan Wang",
"Yuchen Yan",
"Hongxing Li",
"Teng Pan",
"Dingming Li",
"Ruiqing Zhang",
"Weiming Lu",
"Jun Xiao",
"Yueting Zhuang",
"Yongliang Shen"
] | Poster | deep_learning->large_language_models | While long-horizon agentic tasks require language agents to perform dozens of sequential decisions, training such agents with reinforcement learning remains challenging. We identify two root causes: credit misattribution, where correct early actions are penalized due to terminal failures, and sample inefficiency, where... | [
"Large Language Model Agents; Reinforcement Learning; Policy Optimization; Credit Assignment"
] | null | 319 | 2605.06078 | title_snapshot |
hQoT2F37Am | CAOS: Conformal Aggregation of One-Shot Predictors | https://openreview.net/forum?id=hQoT2F37Am | [
"Maja Waldron"
] | Poster | deep_learning->foundation_models | One-shot prediction enables rapid adaptation of pretrained foundation models to new tasks using only one labeled example, but lacks principled uncertainty quantification. While conformal prediction provides finite-sample coverage guarantees, standard split conformal methods are inefficient in the one-shot setting due t... | [
"conformal prediction",
"one-shot prediction"
] | CAOS is a simple conformal aggregation method for one-shot prediction that breaks exchangeability yet retains exact coverage and significantly improves efficiency over split conformal. | 321 | 2601.05219 | title_snapshot |
8nti23Zqkt | EGG: An Expert-Guided Agent Framework for Kernel Generation | https://openreview.net/forum?id=8nti23Zqkt | [
"Yaochen Han",
"Ke Fan",
"Hongxu Jiang",
"Wanqi Xu",
"Weiyu Xie",
"Runhua Zhang",
"Chenhui Zhu",
"Yixiang Zhang"
] | Poster | applications->language_speech_and_dialog | High-performance GPU kernels are critical for reducing the exponentially growing computational costs of large language models (LLMs), but their development heavily relies on manual tuning by domain experts. While recent advances in LLM-based approaches show promise for automating kernel generation, they still struggle ... | [
"Large Language Models",
"Agent",
"Kernel Optimization",
"Code Generation"
] | EGG is an expert-guided agent framework for GPU kernel generation that incorporates expert optimization principles to guide LLMs toward high-performance kernels. | 322 | 2606.26758 | title_snapshot |
0B54C7rFJa | Can We Build a Monolithic Model for Fake Image Detection? SICA: Semantic-Induced Constrained Adaptation for Unified-Yet-Discriminative Artifact Feature Space Reconstruction | https://openreview.net/forum?id=0B54C7rFJa | [
"Bo Du",
"Xiaochen Ma",
"Xuekang Zhu",
"Zhe Yang",
"Chaoqun Niu",
"Jian liu",
"Ji-Zhe Zhou"
] | Poster | applications->computer_vision | Fake Image Detection (FID), aiming at unified detection across four image forensic subdomains, is critical in real-world forensic scenarios. Compared with ensemble approaches, monolithic FID models are theoretically more promising, but to date, consistently yield inferior performance in practice. In this work, we ident... | [
"Fake Image Detection"
] | null | 323 | 2602.06676 | title_snapshot |
iEtOxzAs51 | SGERA: Stein-Guided ECG-Report Alignment for ECG Representation Learning | https://openreview.net/forum?id=iEtOxzAs51 | [
"Jian Chen",
"Yipeng Du",
"Wenhao Yuan",
"Shuai Wang",
"Jinfeng Xu",
"Zewei Liu",
"Running Zhao",
"Edith Cheuk-Han Ngai"
] | Poster | applications->health_medicine | Electrocardiogram (ECG) representation learning via ECG-report alignment is often hindered by the inherent structural and statistical divergence between signals and natural language. Existing methods struggle to bridge this gap with simple contrastive objectives, but struggle with distribution dependencies between hete... | [
"ECG Representation Learning",
"Multi-modal learning",
"Zero-shot classification"
] | This work introduce Stein-Kernel-based alignment instead of CLIP-style into ECG-Report alignment for ECG representation learning, which can deal with distribution dependencies between heterogeneous features. | 328 | null | null |
XdAD8qYsuC | Unsupervised Camouflaged Object Detection with Dual-Eigenvector Spectral Pseudo-Labeling and Contrastive Refinement | https://openreview.net/forum?id=XdAD8qYsuC | [
"Pingzhu Liu",
"Chunming He",
"Zunnan Xu",
"Chao Hao",
"Bo Zhao",
"Xingyu Shao",
"Jun Zhou",
"Zitong YU",
"Xiu Li"
] | Poster | general_machine_learning->unsupervised_and_semisupervised_learning | Unsupervised Camouflaged Object Detection (UCOD) aims to identify objects concealed in their surroundings without relying on pixel-level labels. Existing methods rely solely on simple post-processing of DINO high-dimensional features to generate pseudo labels for training. However, these methods suffer from two major l... | [
"Unsupervised learning",
"camouflaged object detection"
] | This paper is about unsupervised camouflaged object detection. | 334 | null | null |
Ib2IjLO7JW | Adaptive Recurrent Message Passing for Test Time Computing on Graphs | https://openreview.net/forum?id=Ib2IjLO7JW | [
"Junshu Sun",
"Wanxing Chang",
"Qingming Huang",
"Shuhui Wang"
] | Poster | deep_learning->graph_neural_networks | Pre-trained foundation models have demonstrated remarkable success in many domains, enabling a unified backbone to generalize across diverse downstream tasks. However, extending this paradigm to graph learning remains challenging due to the intrinsic mismatch between graph data and fixed architectural designs. In this ... | [
"graph neural networks",
"recurrent model"
] | null | 337 | 2606.22462 | title_snapshot |
BRtDecMvjY | Enhancing LLMs for Graph Tasks via Graph-aware LoRA Generation | https://openreview.net/forum?id=BRtDecMvjY | [
"Junshu Sun",
"Wanxing Chang",
"Qingming Huang",
"Shuhui Wang"
] | Poster | deep_learning->graph_neural_networks | Graph neural networks (GNNs) tightly couple their input-output parameters to dataset-specific feature spaces and target sets, exhibiting limited transferability across different datasets. In contrast, language models (LMs) generalize flexibly via a unified input-output interface, motivating recent attempts to adapt LMs... | [
"graph representation learning",
"large language model",
"low-rank fine-tuning"
] | null | 339 | 2606.22429 | title_snapshot |
ZXm8JjfhW8 | Beyond Attention Imbalance: Mitigating Hallucinations via Spectral Surgery | https://openreview.net/forum?id=ZXm8JjfhW8 | [
"Siqi Lu",
"Wei Suo",
"Yongbin Zheng",
"Jianhang Yao",
"Wanying XU",
"PENG WANG"
] | Poster | deep_learning->large_language_models | While Large Vision-Language Models (LVLMs) achieves remarkable success, hallucinations remain a significant barrier to their reliable deployment. Recent studies primarily attribute these defects to cross-modal attention imbalances, with most solutions focusing on re-weighting visual tokens or suppressing language prior... | [
"Large Vision Language Model",
"Hallucination",
"training-free"
] | null | 358 | null | null |
tSKDiqiCJ4 | Toward Structural Multimodal Representations: Specialization, Selection, and Sparsification via Mixture-of-Experts | https://openreview.net/forum?id=tSKDiqiCJ4 | [
"Hahyeon Choi",
"Nojun Kwak"
] | Poster | general_machine_learning->representation_learning | We propose S3 (Specialization, Selection, Sparsification), a framework that rethinks multimodal learning through a structural perspective. Instead of encoding all signals into a fixed embedding, S3 decomposes multimodal inputs into semantic experts and selectively routes them for each task. Specialization forms concept... | [
"Multimodal Representation Learning",
"Mixture-of-Experts",
"Structured Representation Learning",
"Task-Adaptive Routing"
] | This paper reframes multimodal learning as selecting and pruning semantic components, rather than encoding all information into a single shared embedding. | 361 | 2605.03348 | title_snapshot |
hFocSMm1Ke | Gauge-Equivariant Graph Networks via Self-Interference Cancellation | https://openreview.net/forum?id=hFocSMm1Ke | [
"Yoonhyuk Choi",
"Jiho Choi",
"Jiwoo Kang"
] | Poster | deep_learning->graph_neural_networks | Graph neural networks often degrade on heterophilous graphs because repeated neighbor aggregation can reinforce self-aligned low-frequency components while suppressing phase-inconsistent signals. We propose GESC, a complex-valued graph network that augments attention-based message passing with gauge-consistent U(1) tra... | [
"Graph neural networks",
"graph heterophily",
"self-interference cancellation",
"gauge-equivariant"
] | null | 364 | 2511.16062 | title_snapshot |
vZZUCDcjSx | EchoAttention: Exploiting Token-Pair Redundancy and Frame-Block Similarity for Efficient Video Generation | https://openreview.net/forum?id=vZZUCDcjSx | [
"Yifei Xia",
"Fangcheng Fu",
"Hao Yuan",
"Suhan Ling",
"Xupeng Miao",
"Huixia Li",
"Yuxi Ren",
"Xin Xia",
"Xuefeng Xiao",
"Bin CUI"
] | Poster | deep_learning->attention_mechanisms | Diffusion Transformers (DiTs) are increasingly adopted for long-video generation, yet inference is dominated by the quadratic cost of 3D full attention.
Sparse attention mitigates this bottleneck by exploiting *token-pair redundancy* and pruning query-key interactions.
Nevertheless, its effectiveness on video generati... | [
"Sparse Attention",
"Video Generation",
"Efficient DiT"
] | We propose EchoAttention, combining Sparse attention and Echo frame-block reuse to handle non-sparse heads and speed up attention for video generation. | 367 | null | null |
jciD7uTWu7 | LIFT: A Novel Framework for Enhancing Long-Context Understanding of LLMs via Long Input Fine-Tuning | https://openreview.net/forum?id=jciD7uTWu7 | [
"Yansheng Mao",
"Yufei Xu",
"Jiaqi Li",
"Fanxu Meng",
"Haotong Yang",
"Zilong Zheng",
"Xiyuan Wang",
"Muhan Zhang"
] | Poster | deep_learning->large_language_models | Long-context understanding remains challenging for LLMs due to limited context windows. This paper introduces **Long Input Fine-Tuning (LIFT)**, a framework that improves the long-context performance of arbitrary short-context LLMs by dynamically adapting their parameters to each long input. Instead of endlessly extend... | [
"Large Language Models",
"Long-context Modeling",
"Fine-tuning Strategies",
"Synthetic Data",
"Test-time Tuning"
] | null | 369 | 2502.14644 | title_snapshot |
j3NoRf0n7d | Robust Vision-Language Models via Manifold-Adversarial Adapters | https://openreview.net/forum?id=j3NoRf0n7d | [
"Hao Li",
"Zeyu Xiao",
"Junhao Zhou",
"Peng Liu",
"Yang Zhao",
"Wei Jia"
] | Poster | deep_learning->robustness | Vision-language models (VLMs) have progressed rapidly with large-scale high-quality data and adaptation strategies, yet remain brittle under real-world corruptions, where both visual recognition and language-grounded reasoning degrade. Beyond cascaded image restoration, a natural alternative is parameter-efficient adap... | [
"Large Vision-Language Models",
"Robustness",
"Parameter-Efficient Fine-Tuning",
"Adversarial Training",
"Image Corruptions"
] | We identify the Semantic Misalignment Gap where naive MSE reduction hurts LVLM semantics, and propose Manifold-Adversarial Adapters (MAA) to align corrupted features to the clean manifold for robust reasoning. | 376 | null | null |
DkK7GUr8n3 | ViEEG: Hierarchical Visual Neural Representation for EEG Brain Decoding | https://openreview.net/forum?id=DkK7GUr8n3 | [
"Minxu Liu",
"Donghai Guan",
"Chuhang Zheng",
"Chunwei Tian",
"Jie Wen",
"Qi Zhu"
] | Poster | applications->neuroscience_cognitive_science | Understanding and decoding brain activity into visual representations is a fundamental challenge at the intersection of neuroscience and artificial intelligence. While electroencephalogram (EEG) visual decoding has shown promise due to its non-invasive and low-cost nature, existing methods suffer from {Hierarchical Neu... | [
"Brain-Computer Interfaces",
"Electroencephalogram",
"Neural Representation Learning",
"Cross-modal Alignment",
"Brain Decoding"
] | ViEEG is a biologically inspired hierarchical framework that decodes hierarchical visual information from EEG by aligning brain signals with contour, object, and contextual visual representations. | 386 | 2505.12408 | title_snapshot |
N8JsmN0nQE | OServe: Accelerating LLM Serving via Spatial-Temporal Workload Orchestration | https://openreview.net/forum?id=N8JsmN0nQE | [
"Youhe Jiang",
"Fangcheng Fu",
"Taiyi Wang",
"Guoliang He",
"Eiko Yoneki"
] | Poster | optimization->large_scale_parallel_and_distributed | Serving Large Language Models (LLMs) can benefit immensely from parallelizing both the model and input requests across multiple devices, but incoming workloads exhibit substantial spatial and temporal heterogeneity. Spatially, workloads comprise heterogeneous requests with varying compute and memory demands. Temporally... | [
"LLM Serving",
"Distributed LLM Inference",
"Resource Scheduling",
"Workload Heterogeneity"
] | null | 391 | 2602.12151 | title_snapshot |
jnds7BbY0V | Attention Sink Forges Native MoE in Attention Layers: Sink-Aware Training to Address Head Collapse | https://openreview.net/forum?id=jnds7BbY0V | [
"Zizhuo Fu",
"Wenxuan Zeng",
"Runsheng Wang",
"Meng Li"
] | Poster | deep_learning->attention_mechanisms | Large Language Models (LLMs) often assign disproportionate attention to the first token, a phenomenon known as the attention sink. Several recent approaches aim to address this issue, including Sink Attention in GPT-OSS and Gated Attention in Qwen3-Next. However, a comprehensive analysis of the relationship among these... | [
"Attention Sink",
"Gated Attention",
"Head Collapse"
] | We reveal that attention sink forges native MoE via implicit gating in attention layers, and further propose sink-aware training to address head collapse. | 395 | 2602.01203 | title_snapshot |
Z1HeVZNdMn | Breaking the Synthetic-Real Domain Shortcut for Training-Free Generative Replay-based Class Incremental Learning | https://openreview.net/forum?id=Z1HeVZNdMn | [
"Tao Zhang",
"Qixuan Fan",
"Yiyuan Liang",
"Yanjie Wang",
"Song Yan",
"Tian Tian",
"Jiahuan Zhou",
"Luxin Yan",
"Sheng Zhong",
"Xu Zou"
] | Poster | applications->computer_vision | Class-incremental learning (CIL) requires models to continuously acquire new knowledge while avoiding catastrophic forgetting. While exemplar replay is effective, it raises concerns regarding privacy and storage. Thus, generative replay has emerged as a viable alternative, synthesizing old data using frozen pretrained ... | [
"Class-Incremental Learning",
"Domain Shortcut",
"Feature Rectification"
] | null | 398 | null | null |
KKDi5HuFkt | DiffCrossGait: Trajectory-Level Alignment for 2D-3D Cross-Modal Gait Recognition via Latent Diffusion | https://openreview.net/forum?id=KKDi5HuFkt | [
"Zhiyang Lu",
"Ming Cheng"
] | Poster | applications->computer_vision | Cross-modal 2D–3D gait recognition is impeded by inherent domain discrepancies between 2D silhouette and 3D point cloud distributions. While prior methods align only final embeddings, we propose DiffCrossGait, which enforces trajectory-level alignment by driving both modalities with shared noise in a unified latent dif... | [
"Cross-Modal Learning",
"Diffusion Models",
"Gait Recognition",
"Representation Learning"
] | null | 411 | 2606.00153 | title_snapshot |
ZPJRm2K3oe | ImpText: A Benchmark and Tool-Augmented Framework for Implicit Text Reasoning | https://openreview.net/forum?id=ZPJRm2K3oe | [
"Litao Guo",
"Jinsong Zhou",
"Shuaibo Li",
"Man CHEN",
"Xinli Xu",
"Zixin Zhang",
"Harold Haodong Chen",
"Ying-Cong Chen"
] | Poster | applications->computer_vision | Multimodal Large Language Models (MLLMs) have demonstrated exceptional proficiency in standard text extraction, but they encounter significant challenges when confronting real-world implicit text. Such content typically contains malicious information, intentionally concealed through physical deformation, visual camoufl... | [
"MLLM",
"Benchmark",
"Implicit Text Reasoning"
] | We propose ImpText-Bench and ImpText-Reader to bridge the gap in real-world implicit text recognition. | 412 | null | null |
px1MlO0g26 | Practical Mechanism for Fault-Tolerant Spiking Neural Networks via Simple Input Control Based on Learnable Fragmentation | https://openreview.net/forum?id=px1MlO0g26 | [
"Hyun-Jong Lee",
"Jae-Han Lim"
] | Poster | applications->neuroscience_cognitive_science | Spiking Neural Networks (SNNs) are regarded as the third generation of neural networks, offering energy-efficient computing for neuromorphic devices. Despite this benefit, hardware-implemented SNNs are vulnerable to hardware faults, which severely degrade their performance. Previous approaches have required direct acce... | [
"SNNs",
"hardware faults",
"bottleneck problem",
"practicality",
"learnable fragmentation strategy"
] | We propose a simple fragmentation-based mechanism to enhance the fault tolerance of Spiking Neural Networks (SNNs) in neuromorphic devices. | 421 | null | null |
P8tBsNibfm | Linguistic Relative Policy Optimization for Video Anomaly Reasoning | https://openreview.net/forum?id=P8tBsNibfm | [
"Jiaxu Leng",
"Jiankang Zheng",
"Mengjingcheng Mo",
"Zhanjie Wu",
"Haosheng Chen",
"Ji Gan",
"Xinbo Gao"
] | Poster | applications->computer_vision | Video anomaly detection (VAD) with multimodal large language models has shown strong potential, yet most existing methods still depend on large-scale annotations or expert-designed priors, limiting their ability to acquire anomaly knowledge with as little human intervention as possible. To address this, we propose Ling... | [
"Computer Vision",
"Video Anomaly Reasoning",
"Multimodal Large Language Models",
"Verbalized Learning",
"Reinforcement Learning",
"Tuning-free"
] | null | 424 | null | null |
0dNGyQnDo2 | SceneDirector: Bridging Explicit Geometry and Generative Priors for Unified Driving Scene Editing | https://openreview.net/forum?id=0dNGyQnDo2 | [
"Yiyuan Liang",
"Zhiying Yan",
"Tao Zhang",
"Shangke Liu",
"Kai Lin",
"Xu Zou",
"Nong Sang",
"Changxin Gao"
] | Poster | applications->computer_vision | Validating autonomous driving systems requires diverse scenarios, yet real-world data collection is biased and costly. Editing existing driving logs offers a scalable solution, but simultaneously editing objects and ego-trajectory—termed unified editing—remains challenging.
Current methods face an inherent dilemma: gen... | [
"Diffusion Models",
"Video Editing",
"Autonomous Driving",
"Ego-trajectory Editing",
"Object Editing"
] | SceneDirector bridges explicit geometric guidance with generative priors to enable unified, high-fidelity editing of both objects and ego-trajectories in driving videos within a single inference pass. | 427 | null | null |
ZK0U30zmDY | In-Context Universal Approximation, Compositional Generalization, and Algorithm Emulation | https://openreview.net/forum?id=ZK0U30zmDY | [
"Jerry Yao-Chieh Hu",
"Hong-Yu Chen",
"Po-Chiao Lin",
"Maojiang Su",
"Han Liu"
] | Poster | deep_learning->attention_mechanisms | We study in-context universal approximation and compositional generalization in frozen softmax Transformers as prompt-programmable computation.
We prove in-context universality via in-context emulation:
a fixed-weight Transformer emulates target computations specified by the prompt and hence approximates a broad class... | [
"in-context learning",
"prompt-programmed execution",
"in-context universal approximation",
"in-context algorithm emulation",
"compositional generalization",
"softmax transformers"
] | A constructive theory of in-context universal approximation and function composition by softmax Transformers. | 434 | null | null |
yvDKJfdokC | Learning to Watch: Active Video Anomaly Understanding via Interleaved Policy Optimization | https://openreview.net/forum?id=yvDKJfdokC | [
"Mengjingcheng Mo",
"Jiaxu Leng",
"Xinbo Gao"
] | Poster | applications->computer_vision | Video anomaly understanding (VAU) relies on sparse, context-dependent cues. However, existing passive paradigms suffer from observational aliasing, where static sampling fails to disambiguate semantically distinct events. To overcome this, we propose $Anom\text{-}\pi$, a closed-loop framework that reconceptualizes vide... | [
"Computer Vision",
"Video Anomaly Understanding",
"Large Language Models",
"Reinforcement Learning"
] | Anom-PI formulates video anomaly understanding as active inference with iterative revisiting. | 437 | null | null |
Xxq7fcQUNR | Automated Formal Proofs of Combinatorial Identities via Wilf–Zeilberger Guidance and LLMs | https://openreview.net/forum?id=Xxq7fcQUNR | [
"Beibei Xiong",
"Hangyu Lv",
"Junqi Liu",
"Yisen Wang",
"Shaoshi Chen",
"Jianlin Wang",
"Zhengfeng Yang",
"Lihong Zhi"
] | Spotlight | deep_learning->large_language_models | Automating formal proofs of combinatorial identities is challenging for LLM-based provers, as long-horizon proof planning is required and unconstrained search quickly explodes.
Symbolic methods such as the Wilf--Zeilberger (WZ) method can achieve a mechanized proof of combinatorial identities by constructing special a... | [
"Large Language Models",
"Automated Theorem Proving",
"Combinatoral Identity",
"Wilf–Zeilberger (WZ) Method"
] | null | 439 | 2605.04472 | title_snapshot |
uZ8JZ1Lw9a | Chain-of-Thought Gradient Descent | https://openreview.net/forum?id=uZ8JZ1Lw9a | [
"Hong-Yu Chen",
"Venkat Sripad Ganti",
"Hude Liu",
"Jerry Yao-Chieh Hu",
"Han Liu"
] | Poster | deep_learning->attention_mechanisms | We show that Chain-of-Thought (CoT) expands the expressiveness of Transformer in-context learning (ICL).
Specifically, we show CoT enable efficient simulation of In-Context Gradient Descent (ICGD) for $N$-layer neural network.
Different from CoT, a Transformer with fixed depth and hidden dimension has fixed ICL capaci... | [
"CoT",
"attention"
] | We prove that chain-of-thought with dynamic masking lets a fixed-depth transformer efficiently simulate gradient descent for multi-layer neural networks by reading only the relevant intermediate states. | 443 | null | null |
KD59uBqvNR | Decouple and Cache: KV Cache Construction for Streaming Video Understanding | https://openreview.net/forum?id=KD59uBqvNR | [
"Zhanzhong Pang",
"Dibyadip Chatterjee",
"Fadime Sener",
"Angela Yao"
] | Poster | applications->computer_vision | Streaming video understanding requires processing unbounded video streams with limited memory and computation, posing two key challenges. First, continuously constructing new and evicting old key-value(KV) caches is required for unbounded streams. Secondly, due to the high cost of collecting and training on unbounded s... | [
"streaming video understanding",
"KV Cache"
] | null | 447 | 2605.01858 | title_snapshot |
DKathyl3XN | On Structured State-Space Duality | https://openreview.net/forum?id=DKathyl3XN | [
"Jerry Yao-Chieh Hu",
"Xiwen Zhang",
"Ali ElSheikh",
"Weimin Wu",
"Han Liu"
] | Poster | deep_learning->sequential_models_time_series | Structured State-Space Duality (SSD) [Dao \& Gu, ICML 2024] is an equivalence between a simple Structured State-Space Model (SSM) and a masked attention mechanism.
In particular, a state-space model with a scalar-times-identity state matrix is equivalent to a masked self-attention with a $1$-semiseparable causal mask... | [
"structured state space model",
"attention mechanism"
] | We generalize structured state-space duality to diagonal SSMs and characterize its exact limits for masked attention. | 448 | 2510.04944 | title_snapshot |
xI364qxvhX | CoF-T2I: Video Models as Pure Visual Reasoners for Text-to-Image Generation | https://openreview.net/forum?id=xI364qxvhX | [
"Chengzhuo Tong",
"Mingkun Chang",
"Shenglong Zhang",
"Yuran Wang",
"Cheng Liang",
"Zhizheng Zhao",
"Bohan Zeng",
"Yang Shi",
"Ruichuan An",
"Yifan Dai",
"Ziming Zhao",
"Guanbin Li",
"Pengfei Wan",
"Yuanxing Zhang",
"Wentao Zhang"
] | Poster | deep_learning->generative_models_and_autoencoders | Recent video generation models have revealed the emergence of Chain-of-Frame (CoF) reasoning, enabling frame-by-frame visual inference. With this capability, video models have been successfully applied to various visual tasks (*e.g.*, maze solving, visual puzzles). However, their potential to enhance text-to-image (T2I... | [
"image generation",
"chain-of-frame"
] | Video models as pure visual reasoners for high-quality text-to-image generation via Chain-of-Frame reasoning. | 451 | 2601.10061 | title_snapshot |
qubIJ4i1XQ | GEM: Geometric Entropy Mixing for Optimal LLM Data Curation | https://openreview.net/forum?id=qubIJ4i1XQ | [
"Yue Min",
"Ziyun Qiao",
"Ruining Chen",
"Yujun Li"
] | Poster | deep_learning->large_language_models | LLM pre-training efficacy increasingly depends on data composition rather than sheer volume. Yet, optimal mixing is hindered by categorization flaws: human taxonomies suffer from ontological misalignment, and Euclidean clustering fails to address embedding anisotropy. We introduce **GEM** (**G**eometric **E**ntropy **M... | [
"pretrain",
"data mixing",
"data curation"
] | null | 453 | 2605.26121 | title_snapshot |
kCJ9kaebgd | Instruction Lens Score: Your Instruction Contributes a Powerful Object Hallucination Detector for Multimodal Large Language Models | https://openreview.net/forum?id=kCJ9kaebgd | [
"Runhe Lai",
"Xinhua Lu",
"Yanqi Wu",
"Jinlun Ye",
"Weijiang Yu",
"Ruixuan Wang"
] | Poster | deep_learning->robustness | Multimodal large language models (MLLMs) have achieved remarkable progress, yet the object hallucination remains a critical challenge for reliable deployment. In this paper, we present an in-depth analysis of instruction token embeddings and reveal that they implicitly encode visual information while effectively filter... | [
"Object hallucination detection",
"Large multimodal language model"
] | We propose the InsLen score to detect object hallucination for MLLMs. | 455 | 2605.12258 | title_snapshot |
WVpmVVEA0Y | Deep Ensemble Clustering for Visual Representation Learning | https://openreview.net/forum?id=WVpmVVEA0Y | [
"Yuwei Wang",
"Guikun Chen",
"Xiruo Jiang",
"Yazhou Yao",
"Di Liu",
"Xiangbo Shu",
"Fumin Shen",
"Wenguan Wang"
] | Poster | deep_learning->other_representation_learning | Recent advances in visual representation learning have seen the rise of clustering-based vision backbones, which adopt clustering as a core paradigm for feature extraction. However, existing clustering-based backbones typically rely on a single clustering algorithm, whose inherent inductive bias limits their representa... | [
"Visual Representation Learning",
"Ensemble Clustering",
"Vision Backbone"
] | null | 458 | null | null |
End of preview. Expand in Data Studio
README.md exists but content is empty.
- Downloads last month
- 207