Multimodal Capability Evaluation on MM-Star benchmark leaderboard · SOTA2 Research

Multimodal Capability Evaluation on MM-Star

60.6Average ScoreQwen2.5-VL-7B

Evaluation Results

Method	Links
Qwen2.5-VL-7B2026.03		60.6
Qwen2.5-VL-7B + SFT2026.03		60.4
Qwen2.5-VL-7B + SFT+DPO+GRPO2026.03		60.27
Qwen2.5-VL-7B + SFT+DPO2026.03		59.73
Qwen2.5-VL-7B + SFT+GRPO2026.03		59.33
Qwen2.5-VL-7B + SPA-VL2026.03		59.13
Qwen2.5-VL-7B + SaFeR-VLM2026.03		58.87
Qwen2.5-VL-3B + VLGuard2026.03		52.67
Qwen2.5-VL-7B + TIS2026.03		52.4
LLaVA-NeXT2024.05		51.6
Qwen2.5-VL-3B + SFT2026.03		51.33
Qwen2.5-VL-3B + SFT+GRPO2026.03		51.07
Qwen2.5-VL-3B + SFT+DPO+GRPO2026.03		51
Qwen2.5-VL-3B + SFT+DPO2026.03		50.87
GPT-4V2024.05		50.4
Qwen2.5-VL-7B + VLGuard2026.03		49.87
Qwen2.5-VL-3B2026.03		49.4
Qwen2.5-VL-3B + SPA-VL2026.03		47.2
MiniGemini2024.05		45.5
Qwen2.5-VL-3B + SaFeR-VLM2026.03		44.13
Qwen2.5-VL-3B + TIS2026.03		42.33
OmniLMM + RLAIF-V2024.05		40.9
OmniLMM2024.05		39.7
LLaVA 1.5 + RLAIF-V2024.05		35.4
AMP-MEG2024.05		34.8
Qwen-VL-Chat2024.05		34.5
POVID2024.05		34.3
LLaVA-RLHF2024.05		34.2
VCD2024.05		33.8
Silkie2024.05		33.6
LLaVA 1.52024.05		33.3
RLHF-V2024.05		33.2
Less-is-more2024.05		32.9
OPERA2024.05		32.9
HA-DPO2024.05		32.9
CCA-LLaVA2024.05		32.1