AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

Unified Multimodal Models as Auto-Encoders

Zhiyuan Yan, Kaiqing Lin, Zongjian Li, Junyan Ye, Hui Han, Haochen Wang, Zhendong Wang, Bin Lin, Hao Li, Xinyan Xiao, Jingdong Wang, Haifeng Wang and 1 more

CVPR 2026 · Vision-language and multimodal › Human and embodied modeling · approach: Vision-language pretraining

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
apply Unified-GRPO reconstructive RL post-training (ablation)UniWorld · GenEvaloverall: 84.0 -> 89.0Improved
apply Unified-GRPO (ablation)UniWorld · DPG-Benchoverall: 81.2 -> 86.4Improved
apply Unified-GRPO (ablation)UniWorld · Unified-Scoreunified score: 79.0 -> 86.1Improved
apply Unified-GRPO (ablation)UniWorld · MMMUaccuracy: 58.6 -> 58.2No clear effect
apply Unified-GRPO (ablation)UniWorld · MMBenchaccuracy: 83.5 -> 84.8Improved
apply Unified-GRPO (ablation)Janus-Pro · GenEvaloverall: 80.0 -> 84.3Improved
apply Unified-GRPO (ablation)Janus-Pro · Unified-Scoreunified score: 82.8 -> 89.1Improved
apply Unified-GRPO to Qwen-2.5-VL-3B understanding model (ablation)Qwen-2.5-VL-3B · MMT-Bench overallaccuracy: 56.3 -> 56.5No clear effect
apply Unified-GRPO to Qwen-2.5-VL-3B (ablation)Qwen-2.5-VL-3B · MMT-Bench fine-grained perceptionaccuracy: 32.5 -> 56.9Improved
apply Unified-GRPO to Qwen-2.5-VL-3B (ablation)Qwen-2.5-VL-3B · MMT-Bench Small Object Detectionaccuracy: 5.0 -> 45.0Improved
apply Unified-GRPO to Qwen-2.5-VL-3B (ablation)Qwen-2.5-VL-3B · MMT-Bench Document Understandingaccuracy: 70.0 -> 58.4Hurt
unfreeze visual encoder (ViT) during RL (ablation)Unified-GRPO · reconstruction quality (qualitative)image qualityHurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

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