Saved in:
Bibliographic Details
Main Authors: Wu, Zhixuan, Zha, Quanxing, Wang, Teng, Xu, Genbao, Gu, Wenyuan, Rao, Wei, Ma, Nan, Cheng, Bo, Poria, Soujanya
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.14692
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914568500936704
author Wu, Zhixuan
Zha, Quanxing
Wang, Teng
Xu, Genbao
Gu, Wenyuan
Rao, Wei
Ma, Nan
Cheng, Bo
Poria, Soujanya
author_facet Wu, Zhixuan
Zha, Quanxing
Wang, Teng
Xu, Genbao
Gu, Wenyuan
Rao, Wei
Ma, Nan
Cheng, Bo
Poria, Soujanya
contents Video understanding requires identifying and reasoning over semantically discriminative visual objects across frames, yet existing object-agnostic solutions struggle to effectively handle substantial object variations over time. To address this, we introduce Chain-of-Glimpse, a search-guided progressive object-grounded reasoning framework that explicitly anchors each reasoning step to specific visual evidence regions, enabling compositional and multi-step decision-making. Formally, Chain-of-Glimpse formulates video reasoning as a step-by-step process that incrementally builds spatially grounded traces around task-relevant visual objects, thereby mitigating over-reliance on saliency-driven cues. Specifically, Chain-of-Glimpse features a search-guided controller, optimized via reinforcement learning with a format reward that significantly incentivizes grounding capability, to iteratively ground visual evidence regions and form reliable reasoning trajectories, yielding accurate and interpretable multi-step decisions. Extensive evaluations on both in domain NExTQA and out-of-domain Video-Holmes, CG-Bench Reasoning, and VRBench benchmarks demonstrate consistent performance gains, robustness and generalization of Chain-of-Glimpse across diverse video reasoning tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2604_14692
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Chain-of-Glimpse: Search-Guided Progressive Object-Grounded Reasoning for Video Understanding
Wu, Zhixuan
Zha, Quanxing
Wang, Teng
Xu, Genbao
Gu, Wenyuan
Rao, Wei
Ma, Nan
Cheng, Bo
Poria, Soujanya
Computer Vision and Pattern Recognition
Video understanding requires identifying and reasoning over semantically discriminative visual objects across frames, yet existing object-agnostic solutions struggle to effectively handle substantial object variations over time. To address this, we introduce Chain-of-Glimpse, a search-guided progressive object-grounded reasoning framework that explicitly anchors each reasoning step to specific visual evidence regions, enabling compositional and multi-step decision-making. Formally, Chain-of-Glimpse formulates video reasoning as a step-by-step process that incrementally builds spatially grounded traces around task-relevant visual objects, thereby mitigating over-reliance on saliency-driven cues. Specifically, Chain-of-Glimpse features a search-guided controller, optimized via reinforcement learning with a format reward that significantly incentivizes grounding capability, to iteratively ground visual evidence regions and form reliable reasoning trajectories, yielding accurate and interpretable multi-step decisions. Extensive evaluations on both in domain NExTQA and out-of-domain Video-Holmes, CG-Bench Reasoning, and VRBench benchmarks demonstrate consistent performance gains, robustness and generalization of Chain-of-Glimpse across diverse video reasoning tasks.
title Chain-of-Glimpse: Search-Guided Progressive Object-Grounded Reasoning for Video Understanding
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2604.14692