Share your thoughts, 1 month free Claude Pro on usSee more

Multimodal Reward Modeling on RewardBench Multimodal

99.6Safety Score

R1-Reward

Updated 4d ago

Evaluation Results

Method
R1-Reward 2026.03	99.6	74.9	86.5	58.8	80	-	-	-	-	-	-
MSRL + voting@16 2026.03	96.7	74.3	89.2	65.1	81.3	-	-	-	-	-	-
MSRL 2026.03	96.4	72.8	88.6	64.2	80.5	-	-	-	-	-	-
Gemini-3-Pro 2026.03	94.7	94.2	93	82.4	91.1	-	-	-	-	-	-
Gemini-1.5-Pro 2026.03	94.5	66.3	87.2	63.9	63	-	-	-	-	-	-
MM-RLHF-Reward 2026.03	92.9	54.3	76.8	64.7	72.2	-	-	-	-	-	-
MM-RLHF-Reward 2026.04	92.9	-	-	-	67.1	61.7	67.5	54.3	58.4	57.9	-
InternVL3.5-8B 2026.03	90.6	66.5	83.4	62.6	75.8	-	-	-	-	-	-
IXC-2.5-Reward 2026.04	89.9	-	-	-	66.6	60.7	64.2	56.8	63	50.5	-
Discriminative MRM 2026.03	88.6	68.4	84.2	58.5	74.9	-	-	-	-	-	-
MSRL + voting@16 2026.03	87.2	70.7	84.7	56.8	74.9	-	-	-	-	-	-
MSRL 2026.03	85.4	69.8	84.3	56.2	73.9	-	-	-	-	-	-
Generative MRM 2026.03	84.3	67.6	87.3	65.7	76.2	-	-	-	-	-	-
Qwen3-VL-8B-Instruct 2026.03	82.3	69.7	81.3	60.5	73.5	-	-	-	-	-	-
UnifiedReward-think (w/ CoT) 2026.03	81.5	63.4	78.5	66.4	72.5	-	-	-	-	-	-
Discriminative MRM 2026.03	80.3	64.2	85.8	54.6	71.2	-	-	-	-	-	-
InternVL3.5-14B 2026.03	79.9	73.5	72.4	67.8	73.4	-	-	-	-	-	-
LLaVA-Critic 2026.04	78	-	-	-	54.1	50.1	53.2	51.3	49.2	49.3	-
Generative MRM 2026.03	77.2	65.7	82.8	51.5	69.3	-	-	-	-	-	-
Claude 3.5 Sonnet 2026.04	76.8	-	-	-	71.5	62.6	67.8	73.9	68.6	65.1	-
GPT-4o 2026.03	74.8	72	87.2	65.4	74.9	-	-	-	-	-	-
GPT-4o 2026.04	74.8	-	-	-	70.8	62.6	69	72	67.6	62.1	-
Claude-3.7-Sonnet 2026.03	72	78.1	86.8	66.7	75.9	-	-	-	-	-	-
Claude 3.7 Sonnet 2026.04	72	-	-	-	71.9	58.4	60.7	78.1	76.3	71.3	-
LLaVA-Critic 2026.03	70.7	56	69.4	59.7	64	-	-	-	-	-	-
UnifiedReward 2026.03	64.7	58.8	68.5	56.4	62.1	-	-	-	-	-	-
Skywork-VL-Reward 2026.04	62	-	-	-	65.7	64.5	62.4	54.3	70.8	44.7	-
UnifiedReward-Think 2026.04	55.3	-	-	-	66.7	63.9	68.7	61.1	65.8	57.7	-
VL-MDR 2026.04	51.2	-	-	-	69	69.8	68.7	67.8	78.2	45.2	-
GPT-5-mini 2026.03	50.4	76.8	51.2	70.6	62.3	-	-	-	-	-	-
UnifiedReward 2026.04	50.4	-	-	-	61.6	63.1	58.4	55.9	64.4	44.7	-
GPT-5 2026.04	-	-	-	-	-	-	-	-	-	-	64.6
Claude-Sonnet-4.5 2026.04	-	-	-	-	-	-	-	-	-	-	78.2
Gemini-2.5-Pro 2026.04	-	-	-	-	-	-	-	-	-	-	82.4
InternVL3-8B 2026.04	-	-	-	-	-	-	-	-	-	-	66.9
Qwen2.5-VL-7B 2026.04	-	-	-	-	-	-	-	-	-	-	62.6
Qwen3-VL-4B 2026.04	-	-	-	-	-	-	-	-	-	-	65.9
Qwen3-VL-8B 2026.04	-	-	-	-	-	-	-	-	-	-	71.6
Qwen3-VL-32B 2026.04	-	-	-	-	-	-	-	-	-	-	79
Molmo2-4B 2026.04	-	-	-	-	-	-	-	-	-	-	61.8
Molmo2-8B 2026.04	-	-	-	-	-	-	-	-	-	-	66.8
InternVL3-78B 2026.04	-	-	-	-	-	-	-	-	-	-	75.7
R1-Reward 2026.04	-	-	-	-	-	-	-	-	-	-	82.2
MM-RLHF-Reward 2026.04	-	-	-	-	-	-	-	-	-	-	67.1
LLaVA-Critic 2026.04	-	-	-	-	-	-	-	-	-	-	62.2
Skywork-VL-Reward 2026.04	-	-	-	-	-	-	-	-	-	-	74.2
IXC-2.5-Reward 2026.04	-	-	-	-	-	-	-	-	-	-	66.6
Molmo2-4B Multi-response RM 2026.04	-	-	-	-	-	-	-	-	-	-	73.2
Qwen3-VL-4B Multi-response RM 2026.04	-	-	-	-	-	-	-	-	-	-	71.2