VideoLoop：用循环工作记忆对抗长视频智能体的语义抖动

View PDF HTML (experimental)

Abstract:Long-form video understanding requires multimodal agents to iteratively gather evidence over many reasoning steps. However, most existing agentic methods suffer from semantic thrashing: as append-only working memory grows, attention to key evidence collapses, and the agent loses access to what it has already found. First, we provide a structural argument showing that append-only memory can incorporate newly observed target evidence, but cannot remove accumulated noise or prevent ordered context growth without a rewrite operator. Second, motivated by this analysis, we propose VideoLoop, a multimodal agent with two coupled loops. The outer loop reasons over the video and the inner loop, after each step, retrieves artifacts from an unbounded filesystem of past observations and intermediate analysis, and rewrites a bounded working memory. Extensive experiments demonstrate the effectiveness of VideoLoop, which improves four popular LVLM backbones in a plug-and-play manner, with an average gain of 4.2% points over baseline on VideoMME (long). Further analysis of working memory suggests that VideoLoop mitigates semantic thrashing: on the hardest quarter of VideoMME (long) questions, a blind judge that reads only the agent's context answers 81.1% correctly, versus 60.9% for the append-only agent. With Gemini 3.1 Pro, VideoLoop reaches 88.3% on VideoMME (long), 88.8% on VideoMMMU, and 80.9% on LongVideoBench (long).

Comments:	Code: this https URL
Subjects:	Computer Vision and Pattern Recognition (cs.CV)
Cite as:	arXiv:2609.38119 [cs.CV]
	(or arXiv:2609.38119v2 [cs.CV] for this version)
	https://doi.org/10.48550/arXiv.2609.38119 arXiv-issued DOI via DataCite

Submission history

From: Jinfa Huang [view email]
[v1] Tue, 29 Sep 2026 17:50:21 UTC (564 KB)
[v2] Wed, 30 Sep 2026 03:20:01 UTC (564 KB)