Chain-of-Glimpse: Search-Guided Progressive Object-Grounded Reasoning for Video Understanding
Lists
Only you see this
Abstract
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.
BibTeX
@inproceedings{wu2026chain,
author = {Zhixuan Wu and Quanxing Zha and Teng Wang and Genbao Xu and Wenyuan Gu and Wei Rao and Nan Ma and Bo Cheng and Soujanya Poria},
title = {{Chain-of-Glimpse: Search-Guided Progressive Object-Grounded Reasoning for Video Understanding}},
booktitle = {International Conference on Machine Learning},
year = {2026},
url = {https://openreview.net/forum?id=J9cU4mVavU},
}Loading related content...