AI Paper Atlas

Research map / Video / Video-language understanding

From Evaluation to Enhancement: Benchmarking and Improving Think-with-Video Reasoning for Video Generative Models

Meng Luo, Yicheng Liu, Jiahao Wang, Yuanxing Zhang, Xin Tao, Pengfei Wan, Kun Gai, Hao Fei

ECCV 2026 · Video › Video-language understanding · approach: Event and anomaly reasoning in video · 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 Vid-PRE prompt rewriterWan2.2 · MME-CoFoverall score: 1.72 -> 2.58Improved
add Vid-PRE prompt rewriterWan2.2 · V-ReasonBenchaverage accuracy: 26.65 -> 44.48Improved
add Vid-PRE prompt rewriterWan2.2 · V-ReasonBench spatial cognitionaccuracy: 1.32 -> 36.25Improved
add Vid-PRE prompt rewriterWan2.2 · VWG-Benchoverall score (1-5): 2.13 -> 2.51Improved
add Vid-PRE prompt rewriterWan2.2 · VWG-Benchimplicit rule following: 1.56 -> 2.12Improved
use Vid-PRE instead of Wan native prompt extensionWan2.5 · MME-CoFoverall score: 2.84 -> 3.07Improved
use Vid-PRE instead of Wan native prompt extensionWan2.6-Flash · MME-CoFoverall score: 2.82 -> 3.09Improved
use Vid-PRE instead of Wan native prompt extensionWan2.5 · V-ReasonBenchaverage accuracy: 41.06 -> 51.19Improved
use Vid-PRE instead of Wan native prompt extensionWan2.6-Flash · V-ReasonBenchaverage accuracy: 39.50 -> 52.57Improved
use Vid-PRE instead of Wan native prompt extensionWan2.6-Flash · VWG-Benchoverall score (1-5): 3.04 -> 3.26Improved
add GRPO stage after SFT of Vid-PRE (ablation)Wan2.2 · V-ReasonBenchaverage accuracy: 40.26 -> 44.48Improved
add GRPO stage after SFT of Vid-PRE (ablation)Wan2.6-Flash · V-ReasonBenchaverage accuracy: 45.96 -> 52.57Improved

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