---
title: "COMET：面向视频多模态大模型的时间推理增强框架"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.21030v1"
permalink: "https://feed.astralor.com/items/itm_c36e98e5a5be262d37f379d8"
published: "2026-08-21T12:28:36.000Z"
score: 73
channel: "tech"
tags: ["COMET","Qwen3-VL-8B","TC-GRPO","时间推理","视频多模态大模型"]
---

# COMET：面向视频多模态大模型的时间推理增强框架

## AI 点评

> 这篇论文值得关注，因为它从建模和优化双管齐下，直接戳中视频大模型重空间轻时间的软肋。对做视频理解、Agent时序决策以及端侧视频模型的人，COMET的框架和实验分析都很有参考价值，尤其是不牺牲静态性能这一点，落地性很强。

## 速览

视频多模态大模型在细粒度运动-时间理解上仍然脆弱，核心瓶颈不仅在于帧采样稀疏，更在于缺乏完整的时间建模流程。为此，研究者提出 COMET，一个系统增强视频 MLLM 时间能力的框架。COMET 引入基于泰勒帧差的临时运动分支，通过时间注意力偏置增强的交叉注意力将运动证据注入外观流，并结合时间先验蒸馏与双向 TC-GRPO 优化，将时序方向作为直接学习信号。在 Qwen3-VL-8B 上，动作理解任务平均提升 4.9%，时间推理任务提升 2.1%，且静态感知保持持平；InternVL2.5-8B 上也呈现同样增益，表明具备跨模型泛化能力。

## 关键点

- 提出 COMET 框架，显式建模帧间运动，补齐视频 MLLM 时间建模链路。
- 采用泰勒帧差构建时间运动分支，通过注意力偏置增强的交叉注意力实现外观-运动融合。
- 引入前向-反向 TC-GRPO 优化，将时间顺序作为直接学习信号，并配合时间先验蒸馏。
- 在多个视频理解基准上取得一致提升，且保持静态感知性能不降，证明其通用性。

## 全文

Video multimodal large language models have advanced significantly, yet fine-grained motion-temporal understanding remains fragile. The core bottleneck is not only sparse frame sampling, but also the lack of a complete temporal modeling pipeline for explicitly representing frame-to-frame change, enabling appearance-motion interaction, and optimizing temporal direction sensitivity. We propose COMET, a temporally grounded framework that systematically strengthens video MLLMs through explicit temporal representation, appearance-motion fusion, and direction-aware optimization. Architecturally, COMET introduces a temporal motion branch built on Taylor frame differences and injects its motion evidence into the appearance stream via temporal attention bias-enhanced cross-attention. For optimization, COMET combines temporal prior distillation with a forward-reverse TC-GRPO stage that turns temporal order into a direct learning signal and strengthens the model's use of directional motion patterns encoded by the temporal motion branch. The method achieves consistent overall improvements with a pronounced motion-temporal bias: on Qwen3-VL-8B, action-centric tasks (STAR, SSv2) improve by 4.9% on average, temporal reasoning tasks (NExT-QA, CLEVRER, LLaVA-178K) by 2.1% over BL-GRPO, while static perception tasks (PerceptionTest) remain on par. The same gain pattern also transfers to InternVL2.5-8B, indicating that COMET generalizes across model families.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_c36e98e5a5be262d37f379d8