Multimodal Reward Modeling on MR2Bench Image benchmark leaderboard · SOTA2 Research

Multimodal Reward Modeling on MR2Bench Image

87.1Best-of-4 AccuracyGPT-5

Evaluation Results

Method	Links
GPT-52026.04		87.1
Claude-Sonnet-4.52026.04		72.9
Gemini-2.5-Pro2026.04		71.2
InternVL3-78B2026.04		65
Molmo2-4B Multi-response RM2026.04		62.5
Molmo2-4B2026.04		61.7
Qwen3-VL-4B2026.04		60.8
Qwen3-VL-32B2026.04		60.8
Qwen3-VL-8B2026.04		60.4
Molmo2-8B2026.04		60
R1-Reward2026.04		58.8
Qwen3-VL-4B Multi-response RM2026.04		58.8
LLaVA-Critic2026.04		56.3
InternVL3-8B2026.04		55.4
IXC-2.5-Reward2026.04		55
Skywork-VL-Reward2026.04		52.9
Qwen2.5-VL-7B2026.04		52.5
MM-RLHF-Reward2026.04		45