AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Chimera: Improving Generalist Model with Domain-Specific Experts

Tianshuo Peng, Mingsheng Li, Jiakang Yuan, Hongbin Zhou, Renqiu Xia, Renrui Zhang, Lei Bai, Song Mao, Bin Wang, Aojun Zhou, Botian Shi, Tao Chen and 2 more

ICCV 2025 · Vision-language and multimodal › Multimodal LLMs · approach: Multimodal LLM benchmarks · 10 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 Generalist-Specialist Collaboration Masking at ratio 0.3 (ablation)Chimera-4B · MathVista testminiaccuracy: 59.4 -> 61.3Improved
raise GSCM mask ratio from 0.3 to 0.5 (ablation)Chimera-4B · MathVista testminiaccuracy: 61.3 -> 60.4Hurt
raise GSCM mask ratio from 0.3 to 1.0 (ablation)Chimera-4B · MathVista testminiaccuracy: 61.3 -> 56.2Hurt
add expert encoders without masking instead of naive finetune (ablation)InternVL2-4B · MathVista testminiaccuracy: 58.5 -> 59.4Improved
naively finetune baseline on same data (ablation)InternVL2-4B · MathVista testminiaccuracy: 57.0 -> 58.5Improved
add table and math experts to chart-only expert (ablation)InternVL2-4B · MathVista testminiaccuracy: 59.4 -> 61.3Improved
add Chimera experts to InternVL2-8BInternVL2-8B · MathVista testminiaccuracy: 61.6 -> 64.9Improved
add Chimera experts to InternVL2-2BInternVL2-2B · MathVista testminiaccuracy: 48.3 -> 53.1Improved
add Chimera experts to InternVL2-8BInternVL2-8B · MathVerseaccuracy: 31.3 -> 32.4Improved
apply direct preference optimization on 60K pairsChimera-8B · MathVista testminiaccuracy: 64.9 -> 68.3Improved
add experts to InternVL2-8B in the table domainInternVL2-8B · MathVista table domainaccuracy: 67.1 -> 62.9Hurt
add experts to InternVL2-8B in the math domainInternVL2-8B · MathVista math domainaccuracy: 66.5 -> 71.9Improved

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