Break

Follow this topic Ask about this topic

Momentum

5 papers in the last four weeks, with none the four weeks before. 0.1% of all new papers.

Jul 6Week of Sep 21

Latest papers 34

All topics

Learning Process Rewards via Reasoning State Propagation
Sep 30, 2026Kai Gan, Zi-Hao Zhou, Bo Ye +3Process Reward Model Process-Level Supervision
School of Computer Science and Engineering, Southeast University, Nanjing 210096, China · Key Laboratory of Computer Network and Information Integration (Southeast University), Ministry of Education, China · Zhongguancun Academy +1
Where Do Multi-Agent Systems Fail? Evidence-Grounded Diagnosis of Collective Mechanisms
Sep 30, 2026Zhengye HanDiagnosis Collective Behaviors
New York University New York, NY, USA
Mind the Gap: Mesh-Guided Repair of Broken Vessels
Sep 24, 2026Gniewosz Drwiega, Wojciech Szymanski, Marek WodzinskiVessel Segmentation U-Net
Sano – Centre for Computational Personalised Medicine International Research Foundation, Krakow, Poland · AGH University of Krakow, Krakow, Poland
Reasoning Instructions Can Break Answer Decoding in Vision--Language Models
Sep 24, 2026Zeyan Li, Siyuan Qiu, Jianfeng XuQwen3 Answer
Shanghai Jiao Tong University
Breaking Weather-Content Coupling: Type-Severity Guided Progressive Disentanglement for All-in-One Infrared Restoration
Sep 23, 2026Xinyao Wang, Lijun He, Zhihan Ren +1Query-Specific Degradation-Aware Representation Weather Prediction
Shaanxi Key Laboratory of Deep Space Exploration Intelligent Information Technology, School of Information and Communications Engineering, Xi’an Jiaotong University, Xi’an, 710049, Shaanxi, China
RoleBreak: Benchmarking Long-Horizon Role-Playing Robustness in Spoken Dialogue
Sep 15, 2026Yuqi Wang, Fengyuan Liu, Haochen Luo +2Role-Playing Agents Vocalizations
The University of Hong Kong · Kami AI
What Emerges and What Breaks in Self-Play Driving
Aug 31, 2026Laur Sisask, Ardi Tampuu, Tambet MatiisenSelf-Play Human Driving
University of Tartu, Tartu, Estonia
Role-Break in Attention Heads: Understanding and Detecting Hallucinations in VLMs
Jul 31, 2026Mingyu Wang, Weilin Jin, Wenbo Li +5Large Language Model Hallucination Generative Hallucination
Peking University, Beijing, China · Joy Future Academy, Beijing, China
FORGE-plus: Force-Budgeted Recovery for Contact-Rich Assembly with a Frozen LLM Supervisor
Jul 23, 2026Kyupaeck Jeff Rah, Midum OhGoal-Conditioned Reinforcement Learning Slipsense
Independent Researcher
From Agent Failures to Text Policies: What Works and What Breaks
Jul 22, 2026Jaideep Ray, Ankit GoyalText Generation Text Analysis and Detection
Independent Researcher
Broken Gates: Re-evaluating Web Bot Defenses in the Age of LLM Agents
Jul 21, 2026Behzad Ousat, Nikita Turkmen, Lalchandra Rampersaud +2Captcha Browser Agent
Florida International University
Breaking Déjà Vu: Independent Auditing of Visual Place Recognition through Vision-Language Reasoning
Jul 14, 2026Sania Waheed, Michael Milford, Sarvapali D. Ramchurn +1Visual Place Recognition Robotic Perception
School of Electronics and Computer Science, University of Southampton, Southampton SO17 1BJ, U.K. · School of Electrical Engineering and Computer Science, Queensland University of Technology, Brisbane, QLD 4000, Australia · School of Computer Science and Electronic Engineering, University of Essex, CO4 3SQ Colchester, U.K.
Restore3D: Breathing Life into Broken Objects with Shape and Texture Restoration
Jul 1, 2026Xiaolong Shen, Zongxin Yang, Yi Yang3D Generation Image Restoration
ReLER, CCAI, Zhejiang University, Hangzhou, China
AI Healthcare Chatbots as Information Infrastructure: A Large-Scale Study of User-Reported Breakdowns
Jun 25, 2026Muhammad Hassan, Ramazan Yener, Ece Gumusel +1Healthcare Break
University of Illinois Urbana Champaign, USA · Rutgers University, USA
All Green, Still Broken: Real-Flow Verification Lessons from an LLM-Integrated, Multi-Market Web Application
Jun 21, 2026Muhammad Bilal, Ali Hassaan MughalBrowser Agent Break
Technical University of Munich · Technical University of Munich, Germany · Independent Researcher +1
Trust Between AI Agents: Measuring Formation, Breakage, and Recovery, with Implications for Governing Multi-Agent Systems
Jun 12, 2026Yujiao ChenAI Trustworthiness Artificial Intelligence Agents
Massachusetts Institute of Technology Cambridge, MA 02139
LSTM-Based Detection of Structural Breaks in Property Insurance Loss Reserving: A Climate-Informed Approach
Jun 9, 2026Thomas Mbrice, Shashwat PanigrahiCnn-Lstm Insurance
Stony Brook University · Department of Computer Science
It Takes One to Bias Them All: Breaking Bad with One-Shot GRPO
Jun 9, 2026Naihao Deng, Yilun Zhu, Naichen Shi +2Large Language Model Bias Large Language Model Alignment
University of Michigan · Northwestern University
Breaking Time: A Fully Gaussian Framework for Distributed and Continuous-Time SLAM
Jun 4, 2026Davide Ceriola, Simone Ferrari, Luca Di Giammarino +2Simultaneous Localization And Mapping Gaussian Primitives
Department of Computer, Control, and Management Engineering “Antonio Ruberti”, Sapienza University of Rome, Italy · University of Stuttgart, Germany
Channel Fracture: Three Instances of Cross-Boundary Silent Delivery Reliability Failures in Multi-Agent Systems
Jun 3, 2026Dexing LiuProtocol Break
Shanghai Qijing Digital Technology Co., Ltd.
CaloTrilogy: Toward a Breakthrough in One-Step, End-to-End, Physics-Guided Shower Generation for Modern Calorimeters
Jun 2, 2026Cheng Jiang, Sitian Qian, Kevin Pedro +3Physics-Guided Machine Learning Collision Prediction
School of Physics and Astronomy, University of Edinburgh, Edinburgh EH9 3FD, United Kingdom · Department of Physics, University of Wisconsin-Madison, Madison, WI 53706, USA · Fermi National Accelerator Laboratory, Batavia, IL 60510, USA +2
The Surface You Test Is Not the Surface That Breaks
May 28, 2026Syed Nazmus Sakib, Nafiul Haque, Shahrear Bin Amin +1Attacker Large Language Model Large Language Model Agents
Department of Robotics and Mechatronics Engineering, University of Dhaka · Department of Computer Science and Engineering, University of Dhaka
ReasonBreak: Probing Vulnerabilities in Reasoning-Enabled Vision-Language-Action Models for Autonomous Driving
May 27, 2026Mohammadreza Teymoorianfard, Jean-Philippe Monteuuis, Jonathan Petit +1Autonomous Driving Break
University of Massachusetts Amherst · Qualcomm
StructBreak: Structural Cognitive Overload-Induced Safety Failures in MLLMs
May 25, 2026Yang Luo, Xinran Liu, Tiantian Ji +3Large Language Model Safety Multimodal Large Language Models
Key Laboratory of Trustworthy Distributed Computing and Service (MoE), Beijing University of Posts and Telecommunications · Institute of Computing Technology, Chinese Academy of Sciences
When Bits Break Recourse: Counterfactual-Faithful Quantization
May 16, 2026Chaymae Yahyati, Ismail Lamaakal, Khalid El Makkaoui +1Quantized Lateral String Stability
Mohammed First University
Optimizing Line Segment Inspection with Limited-Range Drones
May 15, 2026José-Miguel Díaz-Báñez, José-Manuel Higes, Alina Kasiuk +1Autonomous Drones Path Planning
Department of Applied Mathematics, University of Seville, Spain
Break-the-Beat! Controllable MIDI-to-Drum Audio Synthesis
May 14, 2026Shuyang Cui, Zhi Zhong, Qiyu Wu +9Break
Sony Group Corporation · Sony AI
A Constraint Programming Approach for n-Day Lookahead Playoff Clinching in the NHL
May 13, 2026Gili Rosenberg, Kyle E. C. Booth, J. Kyle Brubaker +1Competition Break
Amazon Advanced Solutions Lab, Seattle, WA, 98170, USA
Phase Transitions in Affective Meaning Divergence: The Hidden Drift Before the Break
May 9, 2026Napassorn LitchiowongAffective Divergence
School of Computing, National University of Singapore
Breaking Bad: Interpretability-Based Safety Audits of State-of-the-Art LLMs
Apr 22, 2026Krishiv Agarwal, Ramneet Kaur, Colin Samplawski +6Jailbreak Attacks Llm-As-A-Judge
University of Florida · SRI
Breakout-picker: Reducing false positives in deep learning-based borehole breakout characterization from acoustic image logs
Apr 17, 2026Guangyu Wang, Xiaodong Ma, Xinming WuDrilling Data False-Negative Rate
State Key Laboratory of Precision Geodesy, School of Earth and Space Sciences, University of Science and Technology of China, Hefei 230026, China · Mengcheng National Geophysical Observatory, University of Science and Technology of China, Mengcheng 233500, China
Phase Transitions as the Breakdown of Statistical Indistinguishability
Apr 17, 2026Taiyo Narita, Hideyuki MiyaharaPhase Transitions Spatial Photonic Ising Machines
Graduate School of Information Science and Technology, Hokkaido University, Sapporo, Hokkaido 060-0814, Japan
GraphMend: Code Transformations for Fixing Graph Breaks in PyTorch 2
Sep 17, 2025Savini Kashmira, Jayanaka Dantanarayana, Thamirawaran Sathiyalogeswaran +3Pytorch Cpu-Gpu Hybrid Designs
University of Michigan USA · Jaseci Labs USA
An ab initio foundation model of wavefunctions that accurately describes chemical bond breaking
Date pendingAdam Foster, Zeno Schätzle, P. Bernát Szabó +7Quantum Chemistry Dissociation
Microsoft Research AI for Science · Freie Universität Berlin

Sep 30, 2026, cs.AIJ/K move · Enter open · S save

Learning Process Rewards via Reasoning State Propagation

Kai Gan, Zi-Hao Zhou, Bo Ye, Jian Zhao+2

School of Computer Science and Engineering, Southeast University, Nanjing 210096, China · Key Laboratory of Computer Network and Information Integration (Southeast University), Ministry of Education, China · Zhongguancun Academy+1

Process Reward Model Process-Level Supervision Propagation Break

Open paper AskPDF arXiv

Process reward models (PRMs) have demonstrated notable effectiveness in test-time scaling and reinforcement learning by providing fine-grained signals for evaluating intermediate reasoning states, but their training relies heavily on costly process annotations. A natural way to alleviate this dependence is to complement limited process supervision with scalable outcome supervision. However, existing PRMs often model reasoning prefixes independently, providing no explicit mechanism for effectively using final outcome to guide the learning of intermediate reasoning states. We introduce Reasoning State Propagation (RSP), which represents each reasoning prefix with a binary validity state and models transitions between successive states across the reasoning trajectory. Specifically, RSP predicts a break probability that a valid state becomes invalid and a repair probability that an invalid state returns to valid. By propagating these transitions, RSP connects intermediate states to the final state, allowing process annotations to supervise intermediate states while outcome labels supervise the final state and can provide learning signals to preceding steps. Across reasoning search, response selection, and reinforcement learning, RSP consistently outperforms representative PRM baselines, with average improvements over Qwen2.5-Math-PRM of 5.6% in beam search and 2.1% in reinforcement learning.

Momentum

Related topics

Latest papers 34