AI Paper Atlas

Research map / Video / Video-language understanding

Video-KTR: Reinforcing Video Reasoning via Key Token Attribution

Ziyue Wang, Sheng Jin, ZHONGRONG ZUO, Jiawei Wu, Han Qiu, Qi She, Hao Zhang, Xudong Jiang

ICLR 2026 · Video › Video-language understanding · approach: Long video understanding · 14 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 visual-aware token selection (ablation)Vanilla GRPO · Video-Holmes/VideoMMMU/MMVUaverage accuracy: 51.1 -> 52.5Improved
add temporal-aware token selection (ablation)Vanilla GRPO · Video-Holmes/VideoMMMU/MMVUaverage accuracy: 51.1 -> 52.6Improved
add entropy-aware token selection alone (ablation)Vanilla GRPO · Video-Holmes/VideoMMMU/MMVUaverage accuracy: 51.1 -> 51.0No clear effect
add temporal-aware tokens alone (ablation)Vanilla GRPO · VideoMMMUaccuracy: 49.8 -> 50.1Mixed
add entropy tokens to visual and temporal (ablation)V+T · Video-Holmes/VideoMMMU/MMVUaverage accuracy: 52.9 -> 53.4Improved
add temporal tokens to visual and entropy (ablation)V+E · Video-Holmes/VideoMMMU/MMVUaverage accuracy: 52.8 -> 53.4Improved
use hard top-20% token selection instead of soft weighting (ablation)Video-KTR · Video benchmarksaccuracyImproved
use log-probability difference instead of distribution distances (ablation)Video-KTR · Video benchmarksaccuracyImproved
set token update ratio to 20% instead of 30-50% (ablation)Video-KTR · Video benchmarksaccuracyImproved
vary perturbation strategy (mask half, reversal, segmental shuffle) (ablation)Video-KTR · Video benchmarksaverage accuracy: 52.5; 52.6No clear effect
increase input frames from 16 to 64Video-KTR · Video-Holmesaccuracy: 40.7 -> 42.7Improved
replace vanilla GRPO with Video-KTRVanilla GRPO · MMMU(val)accuracy: 44.33 -> 47.33Improved

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