AI Paper Atlas

Research map / Video / Video generation

Building a Precise Video Language with Human-AI Oversight

Zhiqiu Lin, Siyuan Cen, Chancharik Mitra, Isaac Li, Yuhan Huang, Yu Tong Tiffany Ling, Hewei Wang, Irene Pi, Shihang Zhu, Yili Han, Yilun Du, Deva Ramanan

CVPR 2026 · Video › Video generation · approach: Compositional text-to-video diffusion · 1 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 explicit preference and critique signals (full data) to SFT (ablation)Qwen3-VL-8B-Instruct SFT · CHAI benchmark reward modelingaccuracy: 50.9 -> 89.8Improved
add explicit preference and critique signals (full data) to SFT (ablation)Qwen3-VL-8B-Instruct SFT · CHAI benchmark critique generationBLEU-4: 5.5 -> 41.7Improved
add explicit preference and critique signals (full data) to SFT (ablation)Qwen3-VL-8B-Instruct SFT · CHAI benchmark caption generationBLEU-4: 12.0 -> 18.2Improved
add explicit preference and critique signals to DPO (ablation)Qwen3-VL-8B-Instruct DPO · CHAI benchmark caption generationBLEU-4: 9.0 -> 15.8Improved
add explicit preference and critique signals to RLHF-V (ablation)Qwen3-VL-8B-Instruct RLHF-V · CHAI benchmark reward modelingaccuracy: 50.0 -> 81.0Improved
use SFT instead of DPO on full dataQwen3-VL-8B-Instruct · CHAI benchmark caption generationBLEU-4: 15.8 -> 18.2Improved
add second-stage quality control to critiques (ablation)SFT with human critiques · CHAI benchmark reward modelingaccuracy: 73.1 -> 89.8Improved
add second-stage quality control to critiques (ablation)SFT with human critiques · CHAI benchmark critique generationBLEU-4: 35.7 -> 41.7Improved
ensure critique precision (avoid inaccurate critiques) (ablation)SFT with human critiques · CHAI benchmark reward modelingaccuracy: 47.1 -> 89.8Improved
ensure critique recall (avoid incomplete critiques) (ablation)SFT with human critiques · CHAI benchmark reward modelingaccuracy: 56.6 -> 89.8Improved
ensure critique constructiveness (avoid unhelpful critiques) (ablation)SFT with human critiques · CHAI benchmark critique generationBLEU-4: 32.9 -> 41.7Improved
use human expert critiques instead of Gemini-2.5-Pro critiques (ablation)SFT · CHAI benchmark reward modelingaccuracy: 62.0 -> 89.8Improved

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