AI Paper Atlas

Research map / Large language models / Reasoning and chain of thought

Wan-R1: Verifiable-Reinforcement Learning for Generalizable Video Reasoning

Ming Liu, Yunbei Zhang, Shilong Liu, Liwen Wang, Wensheng Zhang

ECCV 2026 · Large language models › Reasoning and chain of thought · approach: RL with verifiable rewards for reasoning

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 Flow-GRPO RL with verifiable rewards after SFTWan-SFT · VR-Bench Regular Mazeexact match: 33.3 -> 61.1Improved
add Flow-GRPO RL with verifiable rewards after SFTWan-SFT · VR-Bench Trapfieldexact match: 38.9 -> 90.3Improved
add Flow-GRPO RL with verifiable rewards after SFTWan-SFT · VR-Bench 3D Mazeexact match: 65.3 -> 94.4Improved
add Flow-GRPO RL with verifiable rewards after SFTWan-SFT · VR-Bench Irregular Mazeexact match: 56.9 -> 47.9Hurt
add RL to SFT model, trained on Easy only, tested on HardWan-SFT · VR-Bench 3D Maze Hardprecision rate: 58.7 -> 96.2Improved
add RL to SFT, Irregular Maze Hard generalizationWan-SFT · VR-Bench Irregular Maze Hardprecision rate: 68.4 -> 51.6Hurt
raise KL coefficient from 0 to 0.1 (ablation)Wan-R1 · VR-Bench Regular Mazestep deviation: 6.7 -> 3.7Improved
raise KL coefficient from 0 to 0.1 (ablation)Wan-R1 · VR-Bench Regular Mazeexact match: 62.5 -> 66.7Improved
reduce training denoising steps from 50 to 30 (ablation)Wan-R1 · VR-Bench Regular Mazeexact match: 62.5 -> 61.1No clear effect
reduce training denoising steps from 50 to 5 (ablation)Wan-R1 · VR-Bench Regular Mazeexact match: 62.5 -> 59.7Hurt
use Qwen2.5-VL as learned reward instead of verifiable reward (ablation)Wan-R1 · VR-Bench mazesreward hackingHurt
use sparse reward only (SR-Only or exact match) (ablation)Wan-R1 · VR-Bench mazesexact matchHurt

Later papers that compared with this method

1 later accepted paper reports Wan-R1 as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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