Research map / Video / Video-language understanding
LVAgent: Long Video Understanding by Multi-Round Dynamical Collaboration of MLLM Agents
Boyu Chen, Zhengrong Yue, Siran Chen, Zikang Wang, Yang Liu, Peng Li, Yali Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Perception retrieval (ablation) | LVAgent without Reflection · LongVideoBench | accuracy: 63.9 -> 69.0 | Improved |
| add Reflection (ablation) | LVAgent without Perception · LongVideoBench | accuracy: 63.9 -> 72.2 | Improved |
| add both Perception and Reflection (ablation) | LVAgent baseline (global 128 frames) · LongVideoBench | accuracy: 63.9 -> 80.0 | Improved |
| add both Perception and Reflection (ablation) | LVAgent baseline (global 128 frames) · VideoMME (w/o subs) | accuracy: 72.1 -> 81.7 | Improved |
| combine three agents instead of one (ablation) | LVAgent, round 1 · LongVideoBench | accuracy: 68.3 -> 69.0 | Improved |
| increase collaboration rounds from 1 to 3 (ablation) | LVAgent · LongVideoBench | accuracy: 69.0 -> 80.0 | Improved |
| increase collaboration rounds from 1 to 3 (ablation) | LVAgent · MLVU | accuracy: 79.2 -> 83.9 | Improved |
| use fine-tuned ASP-CLIP instead of CLIP (ablation) | LVAgent retriever · LongVideoBench | accuracy: 74.7 -> 80.0 | Improved |
| use fine-tuned ASP-CLIP instead of LongCLIP (ablation) | LVAgent retriever · LongVideoBench | accuracy: 75.6 -> 80.0 | Improved |
| use dynamic collaboration instead of Best Score (ablation) | LVAgent discussion · LongVideoBench | accuracy: 76.9 -> 80.0 | Improved |
| use dynamic collaboration instead of Decide by Agent (ablation) | LVAgent discussion · MLVU | accuracy: 79.9 -> 83.9 | Improved |
| raise retrieval threshold from 0.4 to 0.8 (ablation) | LVAgent · LongVideoBench | accuracy: 78.6 -> 80.0 | Improved |
Later papers that compared with this method
4 later accepted papers report LVAgent as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- A Multi-Agent Perception-Action Alliance for Efficient Long Video Reasoning (CVPR 2026 · better on 5 of 5 shared results)
- Symphony: A Cognitively-Inspired Multi-Agent System for Long-Video Understanding (CVPR 2026 · better on 1 of 2 shared results)
- SAGE: Training Smart Any-Horizon Agents for Long Video Reasoning with Reinforcement Learning (CVPR 2026 · better on 1 of 1 shared result)
- Long Story Short: Story-level Video Understanding from 20K Short Films (IJCV 2026 · better on 0 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Towards Effective Long Video Understanding: Dynamic MAS Construction via Meta-Agent (ECCV 2026 · similarity 0.90)
- Symphony: A Cognitively-Inspired Multi-Agent System for Long-Video Understanding (CVPR 2026 · similarity 0.90)
- Vgent: Graph-based Retrieval-Reasoning-Augmented Generation For Long Video Understanding (NeurIPS 2025 · similarity 0.88)
- RTV-Bench: Benchmarking MLLM Continuous Perception, Understanding and Reasoning through Real-Time Video (NeurIPS 2025 · similarity 0.88)
- Searching Videos as Trees: Self-Correcting Agents for Grounded Long Video QA (NeurIPS 2026 · similarity 0.88)
- ReAgent-V: A Reward-Driven Multi-Agent Framework for Video Understanding (NeurIPS 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
