AI Paper Atlas

Research map / Video / Video-language understanding

GTR: Guide-Then-Refine Token Compression for Training-Free Acceleration of Video-LLMs

Sijin Zhou, wei feng, Zhuang Qi, Zhonghua Wang, Lie Ju, Xiang An, Mehrtash Harandi, Zongyuan Ge

ECCV 2026 · Video › Video-language understanding · approach: Video LLMs with fine-grained understanding · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
Add the global score to the text score (ablation)GTR (Slocal-free ablation baseline Stext only) · Average relative performance (LLaVA-OV, 25%)relative performance: 98.3 -> 99.4Improved
Add the local score on top of text and global scores (ablation)GTR · Average relative performance (LLaVA-OV, 25%)relative performance: 99.4 -> 99.7Improved
Add the local score on top of text and global scores (ablation)GTR · MLVUaccuracy: 61.7 -> 61.9Improved
Add the local score on top of text and global scores (ablation)GTR · LongVideoBenchaccuracy: 49.3 -> 49.6Improved
Use text relevance alone instead of local score alone (ablation)GTR single-score variants · MLVUaccuracy: 60.0 -> 61.5Improved
Add GTR to SparseVLMSparseVLM · MLVUrelative performance: 96.3 -> 97.2Improved
Add GTR to SparseVLMSparseVLM · MVBenchrelative performance: 99.1 -> 99.6Improved
Add GTR to FastVFastV · MLVUrelative performance: 94.6 -> 95.4Improved
Use GTR token compression at 25% retention instead of full tokensLLaVA-OV-7B · Total latencytotal latency: 26:03 -> 18:44Improved
Use GTR instead of VidCom2 at 15% retentionLLaVA-OV-7B · Average performanceaccuracy: 54.3 -> 55.7Improved
Reduce retention from 25% to 15%GTR · Average retained performance (LLaVA-OV)relative performance: 99.8 -> 95.8Hurt
Use GTR instead of DyCoke on Qwen2-VLQwen2-VL · VideoMME 25% retentionrelative performance: 89.5 -> 94.5Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy