List your company
ResearchFormerly WizWand

Reward Modeling on RoboRewardBench

0.665MAERoboReward (8B)
0.627320.881661.1361.39034Jan 2, 2026Feb 1, 2026Mar 4, 2026Apr 4, 2026May 5, 2026Jun 5, 2026Jul 6, 2026
Updated 3mo ago

Evaluation results

MethodLinks
RoboReward (8B)Parameters=8B, Access=Open2026.010.665
GPT-5 mini (2025-08-07)Access=Limited2026.010.691
LLM-as-a-VerifierBase model=RoboReward 8B, Reward formulation=Continuous, K=82026.070.72
GPT-5 (2025-08-07)Access=Limited2026.010.811
RoboReward (4B)Parameters=4B, Access=Open2026.010.845
Gemini 3 Pro (Preview)Access=Limited2026.010.851
GPT-5.2 (2025-12-11)Access=Limited2026.010.887
Qwen3-VL Instruct (8B)Parameters=8B, Access=Open2026.010.892
GPT-5.1 (2025-11-13)Access=Limited2026.010.901
Gemini 2.5 ProAccess=Limited2026.010.902
Qwen3-VL Instruct (30B)Parameters=30B, Access=Open2026.010.903
Gemini Robotics-ER 1.5Access=Limited2026.010.906
Gemini 3 Flash (Preview)Access=Limited2026.010.917
Gemini 2.5 FlashAccess=Limited2026.010.943
Gemini 2.5 Flash-LiteAccess=Limited2026.010.99
Qwen2.5-VL Instruct (72B)Parameters=72B, Access=Open2026.010.991
Qwen3-VL Instruct (4B)Parameters=4B, Access=Open2026.011.032
RoboReward 8BReward formulation=Discrete2026.071.11
Qwen2.5-VL Instruct (32B)Parameters=32B, Access=Open2026.011.137
Qwen2.5-VL Instruct (7B)Parameters=7B, Access=Open2026.011.172
Llama 4 Maverick InstructAccess=Open2026.011.271
GPT-5 nano (2025-08-07)Access=Limited2026.011.295
Llama 4 Scout InstructAccess=Open2026.011.485
Qwen2.5-VL Instruct (3B)Parameters=3B, Access=Open2026.011.607