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

Reward Modeling on RoboRewardBench

0.665MAE

RoboReward (8B)

Updated 1mo ago

Evaluation Results

Method	Links
RoboReward (8B) 2026.01		0.665
GPT-5 mini (2025-08-07) 2026.01		0.691
GPT-5 (2025-08-07) 2026.01		0.811
RoboReward (4B) 2026.01		0.845
Gemini 3 Pro (Preview) 2026.01		0.851
GPT-5.2 (2025-12-11) 2026.01		0.887
Qwen3-VL Instruct (8B) 2026.01		0.892
GPT-5.1 (2025-11-13) 2026.01		0.901
Gemini 2.5 Pro 2026.01		0.902
Qwen3-VL Instruct (30B) 2026.01		0.903
Gemini Robotics-ER 1.5 2026.01		0.906
Gemini 3 Flash (Preview) 2026.01		0.917
Gemini 2.5 Flash 2026.01		0.943
Gemini 2.5 Flash-Lite 2026.01		0.99
Qwen2.5-VL Instruct (72B) 2026.01		0.991
Qwen3-VL Instruct (4B) 2026.01		1.032
Qwen2.5-VL Instruct (32B) 2026.01		1.137
Qwen2.5-VL Instruct (7B) 2026.01		1.172
Llama 4 Maverick Instruct 2026.01		1.271
GPT-5 nano (2025-08-07) 2026.01		1.295
Llama 4 Scout Instruct 2026.01		1.485
Qwen2.5-VL Instruct (3B) 2026.01		1.607