The year/Independent research

Paper 2512.20618

LongVideoAgent: Multi-Agent Reasoning with Long Videos

Published
Dec 2025
Research lab
Independent
Citations
21
GitHub
126 stars

01 In brief

Summary

LongVideoAgent is a multi-agent framework for long-video question answering.

A master LLM coordinates a grounding agent to localize question-relevant segments and a vision agent to extract targeted visual observations.

The master agent plans with a step limit and is trained with reinforcement learning (GRPO) to encourage concise, correct, and efficient cooperation.

The system outperforms non-agent baselines on the proposed LongTVQA and LongTVQA+ datasets, which aggregate TVQA/TVQA+ clips into episode-level sequences.

Ablations show that grounding and vision are essential, increasing the step limit from 2 to 5 improves accuracy, larger evidence windows help, and stronger vision models yield higher accuracy.

Reinforcement learning further boosts performance, especially for smaller open-source models.

02 From the paper

Abstract

Recent advances in multimodal LLMs and systems that use tools for long-video QA point to the promise of reasoning over hour-long episodes. However, many methods still compress content into lossy summaries or rely on limited toolsets, weakening temporal grounding and missing fine-grained cues. We propose a multi-agent framework in which a master LLM coordinates a grounding agent to localize question-relevant segments and a vision agent to extract targeted textual observations. The master agent plans with a step limit, and is trained with reinforcement learning to encourage concise, correct, and efficient multi-agent cooperation. This design helps the master agent focus on relevant clips via grounding, complements subtitles with visual detail, and yields interpretable trajectories. On our proposed LongTVQA and LongTVQA+ which are episode-level datasets aggregated from TVQA/TVQA+, our multi-agent system significantly outperforms strong non-agent baselines. Experiments also show reinforcement learning further strengthens reasoning and planning for the trained agent. Code and data will be shared at https://longvideoagent.github.io/.