Salvato in:
Dettagli Bibliografici
Autori principali: Lee, Daeun, Yu, Shoubin, Zhang, Yue, Bansal, Mohit
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:https://arxiv.org/abs/2603.14659
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912967471136768
author Lee, Daeun
Yu, Shoubin
Zhang, Yue
Bansal, Mohit
author_facet Lee, Daeun
Yu, Shoubin
Zhang, Yue
Bansal, Mohit
contents Video reasoning requires models to locate and track question-relevant evidence across frames. While reinforcement learning (RL) with verifiable rewards improves accuracy, it still struggles to achieve reliable spatio-temporal grounding during the reasoning process. Moreover, improving grounding typically relies on scaled training data or inference-time perception tools, which increases annotation cost or computational cost. To address this challenge, we propose VisonCoach, an input-adaptive RL framework that improves spatio-temporal grounding through visual prompting as training-time guidance. During RL training, visual prompts are selectively applied to challenging inputs to amplify question-relevant evidence and suppress distractors. The model then internalizes these improvements through self-distillation, enabling grounded reasoning directly on raw videos without visual prompting at inference. VisonCoach consists of two components: (1) Visual Prompt Selector, which predicts appropriate prompt types conditioned on the video and question, and (2) Spatio-Temporal Reasoner, optimized with RL under visual prompt guidance and object-aware grounding rewards that enforce object identity consistency and multi-region bounding-box overlap. Extensive experiments demonstrate that VisonCoach achieves state-of-the-art performance under comparable settings, across diverse video reasoning, video understanding, and temporal grounding benchmarks (V-STAR, VideoMME, World-Sense, VideoMMMU, PerceptionTest, and Charades-STA), while maintaining a single efficient inference pathway without external tools. Our results show that visual prompting during training improves grounded video reasoning, while self-distillation enables the model to internalize this ability without requiring prompts at inference time.
format Preprint
id arxiv_https___arxiv_org_abs_2603_14659
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle VisionCoach: Reinforcing Grounded Video Reasoning via Visual-Perception Prompting
Lee, Daeun
Yu, Shoubin
Zhang, Yue
Bansal, Mohit
Computer Vision and Pattern Recognition
Artificial Intelligence
Video reasoning requires models to locate and track question-relevant evidence across frames. While reinforcement learning (RL) with verifiable rewards improves accuracy, it still struggles to achieve reliable spatio-temporal grounding during the reasoning process. Moreover, improving grounding typically relies on scaled training data or inference-time perception tools, which increases annotation cost or computational cost. To address this challenge, we propose VisonCoach, an input-adaptive RL framework that improves spatio-temporal grounding through visual prompting as training-time guidance. During RL training, visual prompts are selectively applied to challenging inputs to amplify question-relevant evidence and suppress distractors. The model then internalizes these improvements through self-distillation, enabling grounded reasoning directly on raw videos without visual prompting at inference. VisonCoach consists of two components: (1) Visual Prompt Selector, which predicts appropriate prompt types conditioned on the video and question, and (2) Spatio-Temporal Reasoner, optimized with RL under visual prompt guidance and object-aware grounding rewards that enforce object identity consistency and multi-region bounding-box overlap. Extensive experiments demonstrate that VisonCoach achieves state-of-the-art performance under comparable settings, across diverse video reasoning, video understanding, and temporal grounding benchmarks (V-STAR, VideoMME, World-Sense, VideoMMMU, PerceptionTest, and Charades-STA), while maintaining a single efficient inference pathway without external tools. Our results show that visual prompting during training improves grounded video reasoning, while self-distillation enables the model to internalize this ability without requiring prompts at inference time.
title VisionCoach: Reinforcing Grounded Video Reasoning via Visual-Perception Prompting
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2603.14659