arXiv AI

Stepwise Intrinsic Rewards for Reasoning in Large Language Models

The paper introduces Stepwise Marginal Information Gain (MIG), an intrinsic process reward that evaluates how each reasoning step of a large language model (LLM) or vision-language model (VLM) improves the likelihood of the reference answer. MIG rewards only new likelihood maxima, preventing duplicate credit, and is combined with outcome, format, and self‑distillation objectives to guide training. Experiments on eight benchmarks show that this method outperforms outcome‑only reinforcement learning and improves accuracy by up to 4.8 points over binary‑reward training, including a 12.6‑point gain on MathVerse and a 12.9‑point advantage on vision‑language transfer at 7B parameters.

arXiv Computer Vision
Oct 23

RECAP: Looking Once Is Not Enough for Vision-Language Reasoning

The paper introduces RECAP, a training method that enhances vision‑language models by incorporating temporally localized reactivation of image evidence—called visual revisit—into the reinforcement learning process. Visual revisit peaks are shown to predict correctness and causally improve performance; masking them hurts accuracy more than masking non‑peaks, while adding revisit peaks to failed reasoning partially restores correct answers. RECAP applies revisit‑conditioned advantage estimation, dynamic discounting, and gated value prediction to propagate final rewards through visually important reasoning steps, achieving significant gains on HallusionBench and other benchmarks without extra rewards or inference‑time changes.

By Zhaolu Kang, Tailong Luo, Chenxin Li, Yingjie He, Zhenyu Yu, Fengyu Zhou, Jiachen Qian, Lei Wei, Shuang Chen, Jiachen Li, Eric Hanchen Jiang, Rongchao Zhang, Zhengtao Yao, Hoi Leong Lee, Guansu Wang, Kaiyue Zhou
arXiv Computer Vision
Oct 12

OWLET: On-Policy Weighted Learning via Evidence Transfer for Video Large Language Model Post-Training

The paper introduces OWLET, an on-policy weighted learning method for post-training Video Large Language Models. OWLET leverages a richer teacher view that combines global temporal context with question-conditioned local windows to provide visual evidence, and it employs evidence-aware credit redistribution and spatiotemporal state alignment to better utilize this evidence during reinforcement learning. The approach aims to improve video reasoning by aligning token-level supervision with spatiotemporal representations.

By Qitan Lv, Ke Fan, Hong Wang, Wen Wu, Yilin Wang, Bowen Zhou, Feng Wu, Chao Zhang
arXiv AI
Jul 17

Stop Thinking, Start Looking: Efficient Post-Training for Multimodal Document Question Answering via Reasoning-Free Alignment

arXiv:2607. 14682v1 Announce Type: new Abstract: Efficient multimodal document question answering with explicit visual grounding, locating the precise document region that supports each answer remains an open challenge.

By Harikrishnan P M, Goutham Vignesh, Ganesh Parab, Saisubramaniam Gopalakrishnan, Vishal Vaddina, Varun V, Rohit Agrawal
Hugging Face Trending Papers
Aug 8

StructReward: Efficient Structured Process Rewards for Self-Correcting Multimodal Reasoning

Reinforcement learning with verifiable rewards (RLVR) has emerged as an effective approach for improving multimodal reasoning. However, most existing methods evaluate an entire response using a binary reward based only on final-answer correctness, thereby discarding the supervision available in intermediate reasoning steps.

arXiv Machine Learning
Jul 7

Incentivizing Vision Language Models to Search for Long Video Question Answering

arXiv:2607. 02959v1 Announce Type: cross Abstract: We introduce VSeek, an agentic framework that transforms long-video question answering (LVQA) from a passive, single-pass perception task into a multi-turn retrieval process.

By Harsh Goel, S P Sharan, Sahil Shah, Minkyu Choi, Joungbin An, Kristen Grauman, Sandeep P. Chinchali
arXiv AI
Oct 2

Your Language Model is Its Own Critic: Reinforcement Learning with Value Estimation from Actor's Internal States

The paper introduces POISE, a reinforcement learning algorithm that uses a model’s internal states as a value estimator to reduce variance in reinforcement learning with verifiable rewards (RLVR). By employing a lightweight probe that reads internal signals during the forward pass, POISE predicts baselines online and uses a cross‑rollout construction to keep gradients unbiased. Experiments on Qwen3‑4B and OLMo3‑7B‑Instruct‑DPO across six domains show POISE outperforms existing RLVR baselines, offering more stable training and a value model that generalizes across tasks and scales with the policy.

By Yunho Choi, Jongwon Lim, Woojin Ahn, Minjae Oh, Jeonghoon Shim, Yohan Jo
arXiv Computer Vision
Oct 16

Perception or Reasoning? Diagnosing and Tackling Reinforcement Learning Bottlenecks in VLMs

The paper introduces Perception and Reasoning Attribution (PARA), a method that quantifies how much of a vision‑language model’s errors can be fixed by better perception versus stronger reasoning. PARA reveals that reinforcement learning (RL) benefits more from improving reasoning than perception, and motivates a new group‑relative attention distillation objective called GRAD. GRAD directly guides weaker rollouts toward the visual attention patterns of the best rollouts, consistently boosting performance across multiple model scales and benchmarks.

By Zeyuan Yang, Zhuohao Chen, Jiaben Chen, Yao Zhang, Maohao Shen, Chuang Gan
arXiv AI
Jun 8

Teaching the Way, Not the Answer: Privileged Tutoring Distillation for Multimodal Policy Optimization

arXiv:2606. 07000v1 Announce Type: new Abstract: Recent post-training methods, particularly Reinforcement Learning with Verifiable Rewards (RLVR), have significantly enhanced the reasoning ability of Large Vision-Language Models (LVLMs).

By Shizhe Xiang, Ke An, Wenlong Yu, Yue Liu, Jian Luan, Pei Fu, Qilong Wang