AI Paper Atlas

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

OneCAT: Decoder-Only Auto-Regressive Model for Unified Understanding and Generation

Han Li, Xinyu Peng, Yaoming Wang, Zelin Peng, Xin Chen, Rongxiang Weng, Jingang Wang, Xunliang Cai, Wenrui Dai, Hongkai Xiong

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
add distillation of all layers' hidden states from the teacher (ablation)OneCAT-1.5B understanding stage · 7-benchmark average (8B tokens)average score: 31.4 -> 35.3Improved
distill all layers instead of only the last layer's hidden states (ablation)OneCAT-1.5B understanding stage · 7-benchmark averageaverage score: 33.9 -> 35.3Improved
use full-layer distillation instead of EvE or VoRA distillation (ablation)OneCAT-1.5B understanding stage · 7-benchmark averageaverage score: 33.4 (EvE), 32.3 (VoRA) -> 35.3Improved
use custom teacher (InternViT + same Qwen2.5 LLM) instead of Qwen2.5-VL teacher (ablation)OneCAT-1.5B distillation · 7-benchmark averageaverage score: 33.7 -> 35.3Improved
distill both visual and text tokens instead of visual tokens only (ablation)OneCAT-1.5B distillation · 7-benchmark averageaverage score: 34.8 -> 35.3Improved
add scale-aware adapter (SAA) (ablation)OneCAT-1.5B stage-2 · GenEvalGenEval: 78.1 -> 81.2Improved
add scale-aware adapter (SAA) (ablation)OneCAT-1.5B stage-2 · DPG-BenchDPG: 74.0 -> 74.9Improved
increase generation tokens in mid-training from 10B to 30B (ablation)OneCAT-1.5B stage-2 · GenEvalGenEval: 80.7 -> 81.2Improved
increase generation tokens in mid-training (ablation)OneCAT-1.5B stage-2 · Understanding averageaverage score: 55.1 -> 56.0 (0B -> 30B), 55.2 at 45BNo clear effect
early-fusion decoder-only instead of encoder-based late fusion (no distillation) (ablation)OneCAT-1.5B stage-1 · 7-benchmark average (70B tokens)average score: 37.8 -> 38.3No clear effect
remove vision encoder, use decoder-only multi-scale AR generation vs diffusion BAGELOneCAT-3B · 1024x1024 T2I inferenceinference time (s): 26.29 -> 2.85Improved
remove external ViT for understanding prefillingOneCAT-3B vs Qwen2.5-VL-3B · 1792x1792 inputTTFT (s): 0.583 -> 0.225Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 2 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