List your company

Multimodal Chain-of-Thought Reasoning on M3CoT

79.7AccuracyMMS-PRM
29.88442.81755.7568.683Apr 7, 2026Apr 16, 2026Apr 26, 2026May 6, 2026May 16, 2026May 26, 2026Jun 5, 2026

Evaluation results

MethodLinks
MMS-PRMSize=8B2026.0679.773
InternVL2.5-MPO + SFTSize=8B2026.0675.771
InternVL2.5-MPOSize=8B2026.0667.569.4
LLaVA-NeXT-LLaMA3 + SFTSize=8B2026.0667.460.8
InternVL-2.5Size=8B2026.0662.468
Insight-VSize=7B2026.0661.565.7
Qwen2-VLSize=7B2026.0657.865.1
MiniCPM-V-2-6Size=8B2026.065664.2
LLaVA-OneVisionSize=7B2026.0652.364.6
Ours (PSP)Model=LaViDa, Generation length L / step T=256/1282026.0450.5—
Ours (PSP & VRG)Model=LaViDa, Generation length L / step T=256/1282026.0450.3—
MarginModel=LaViDa, Generation length L / step T=256/1282026.0449.2—
Low-confModel=LaViDa, Generation length L / step T=256/1282026.0449—
Ours (PSP & VRG)Model=LaViDa, Generation length L / step T=128/642026.0448.6—
Ours (PSP & VRG)Model=LaViDa, Generation length L / step T=64/322026.0448.4—
CCoTModel=LaViDa, Generation length L / step T=256/1282026.0447.7—
Ours (PSP)Model=LaViDa, Generation length L / step T=64/322026.0447.6—
Ours (PSP)Model=LaViDa, Generation length L / step T=128/642026.0447.3—
EntropyModel=LaViDa, Generation length L / step T=128/642026.0446.9—
EntropyModel=LaViDa, Generation length L / step T=256/1282026.0446.8—
DDCoTModel=LaViDa, Generation length L / step T=128/642026.0446.7—
DDCoTModel=LaViDa, Generation length L / step T=256/1282026.0446.7—
CCoTModel=LaViDa, Generation length L / step T=128/642026.0446.5—
MarginModel=LaViDa, Generation length L / step T=128/642026.0446.5—
EntropyModel=LaViDa, Generation length L / step T=64/322026.0446.4—
MarginModel=LaViDa, Generation length L / step T=64/322026.0446.3—
Low-confModel=LaViDa, Generation length L / step T=128/642026.0446.2—
Low-confModel=LaViDa, Generation length L / step T=64/322026.0445.8—
DDCoTModel=LaViDa, Generation length L / step T=64/322026.0445.7—
LLaVA-NeXT-LLaMA3Size=8B2026.0645.652.1
CCoTModel=LaViDa, Generation length L / step T=64/322026.0445.2—
LLaVA-1.5Size=13B2026.0639.5—
Ours (PSP & VRG)Model=MMaDa, Generation length L / step T=128/642026.0436.6—
Ours (PSP & VRG)Model=MMaDa, Generation length L / step T=256/1282026.0436.4—
Ours (PSP & VRG)Model=MMaDa, Generation length L / step T=64/322026.0436.3—
Ours (PSP)Model=MMaDa, Generation length L / step T=256/1282026.0435.8—
Ours (PSP)Model=MMaDa, Generation length L / step T=64/322026.0435.6—
Ours (PSP)Model=MMaDa, Generation length L / step T=128/642026.0435.2—
CCoTModel=MMaDa, Generation length L / step T=64/322026.0434.7—
Low-confModel=MMaDa, Generation length L / step T=256/1282026.0434.6—
MarginModel=MMaDa, Generation length L / step T=64/322026.0434.5—
EntropyModel=MMaDa, Generation length L / step T=256/1282026.0434.3—
DDCoTModel=MMaDa, Generation length L / step T=64/322026.0434.1—
DDCoTModel=MMaDa, Generation length L / step T=256/1282026.0434.1—
EntropyModel=MMaDa, Generation length L / step T=64/322026.0434.1—
DDCoTModel=MMaDa, Generation length L / step T=128/642026.0434—
MarginModel=MMaDa, Generation length L / step T=128/642026.0433.8—
MarginModel=MMaDa, Generation length L / step T=256/1282026.0433.8—
Low-confModel=MMaDa, Generation length L / step T=128/642026.0433.8—
Low-confModel=MMaDa, Generation length L / step T=64/322026.0433.7—
EntropyModel=MMaDa, Generation length L / step T=128/642026.0433.6—
CCoTModel=MMaDa, Generation length L / step T=256/1282026.0433.3—
CCoTModel=MMaDa, Generation length L / step T=128/642026.0431.8—