List your company
ResearchFormerly WizWand

Reward Modeling on RMB

89.3AccuracySkywork-Reward-V2-Llama-3.1-8B-40M
63.61270.28176.9583.619May 5, 2025Jun 30, 2025Aug 25, 2025Oct 21, 2025Dec 16, 2025Feb 10, 2026Apr 8, 2026
Updated 6mo ago

Evaluation results

MethodLinks
Skywork-Reward-V2-Llama-3.1-8B-40MParadigm=Our Reward Models2025.0789.3
Skywork-Reward-V2-Llama-3.1-8BParadigm=Our Reward Models2025.0786.4
DeepSeek-V3.22026.0383.9
Skywork-Reward-V2Backbone=gemma-2-9b-it2025.0783.9
FARE-8BStage=SFT, Data=2.5M2026.0383.2
Skywork-Reward-V2-Qwen3-8BParadigm=Our Reward Models2025.0781.2
Skywork-Reward-V2-Qwen3-4BParadigm=Our Reward Models2025.0780.6
Skywork-Reward-V2-Llama-3.2-3BParadigm=Our Reward Models2025.0780.5
Mix-GRMStage=SFT, RL, Data=9K, 21K2026.0380.1
Gemini-3-Flash2026.0379.2
Base-GRMStage=SFT, Data=9K2026.0379.2
ConsistRM (Qwen3-8B)GT=False2026.0479.1
Mix-GRMStage=SFT, Data=9K2026.0378.9
Qwen3-8B2026.0478.8
RubricRM-8BStage=SFT, Data=36K2026.0378.5
Base-GRMStage=SFT, RL, Data=9K, 21K2026.0378.5
Skywork-Reward-V2Backbone=Qwen2.5-7B2025.0778.3
Qwen3-8B + RFTGT=True2026.0478.2
Skywork-Reward-V2-Qwen3-1.7BParadigm=Our Reward Models2025.0778.1
Qwen3-8B + CAARGT=False2026.0478.1
Qwen3-4B + CAARGT=False2026.0477.5
Skywork-Reward-V2Backbone=gemma-3-1b-it2025.0777.1
Skywork-Reward-V2Backbone=gemma-3-4b2025.0777.1
Qwen3-8B + SFTGT=True2026.0476.9
Skywork-Reward-V2-Llama-3.2-1BParadigm=Our Reward Models2025.0776.7
Skywork-Reward-V2Backbone=gemma-2-2b-it2025.0776.7
Qwen3-4B + RFTGT=True2026.0476.7
Qwen3-4B + SFTGT=True2026.0476.5
ConsistRM (Qwen3-4B)GT=False2026.0476.5
Qwen3-4B2026.0476.4
OPRM-Qwen2.5-32BModel=OPRM, Backbone=Qwen2.5-32B, Region Flooding Tuning (RgFT)=false, Source (Reported vs. Reproduced)=Ours2026.0275.6
Qwen3-4B + TTRLGT=False2026.0475.6
OPRM-RgFT-Qwen2.5-32BModel=OPRM, Backbone=Qwen2.5-32B, Region Flooding Tuning (RgFT)=true, Source (Reported vs. Reproduced)=Ours2026.0274.8
Skywork-Reward-V2-Qwen3-0.6BParadigm=Our Reward Models2025.0774.5
Skywork-Reward-8BStage=BT, Data=44K2026.0374.4
OPRM-RgFT-Qwen2.5-72BModel=OPRM, Backbone=Qwen2.5-72B, Region Flooding Tuning (RgFT)=true, Source (Reported vs. Reproduced)=Ours2026.0274.2
Qwen3-8B + TTRLGT=False2026.0474.2
GPT-4oModel=GPT-4o, Source (Reported vs. Reproduced)=Reported2026.0273.8
OPRM-Qwen2.5-14BModel=OPRM, Backbone=Qwen2.5-14B, Region Flooding Tuning (RgFT)=false, Source (Reported vs. Reproduced)=Ours2026.0273.8
GPT-4oParadigm=LLM-as-a-Judge & Generative Reward Models2025.0773.8
GPT-4o-0806Reward Model Category=GenRMs2025.0573.8
OPRM-Qwen2.5-72BModel=OPRM, Backbone=Qwen2.5-72B, Region Flooding Tuning (RgFT)=false, Source (Reported vs. Reproduced)=Ours2026.0273.5
JudgeLRM-7BStage=RL, Data=100K2026.0373.1
RM-R1-Qwen-Instruct-32BParadigm=LLM-as-a-Judge & Generative Reward Models2025.0773
RM-R1-QWEN-INSTRUCT-32BReward Model Category=Our Methods2025.0573
OPRM-RgFT-Qwen2.5-14BModel=OPRM, Backbone=Qwen2.5-14B, Region Flooding Tuning (RgFT)=true, Source (Reported vs. Reproduced)=Ours2026.0272.8
Starling-RM-34BParadigm=Bradley-Terry2025.0772
OPRM-Qwen2.5-7BModel=OPRM, Backbone=Qwen2.5-7B, Region Flooding Tuning (RgFT)=false, Source (Reported vs. Reproduced)=Ours2026.0271.5
DeepSeek-GRM-16BStage=SFT, RL, Data=1.2M, 237K2026.0370.8
Claude-3.5-sonnetModel=Claude-3.5-sonnet, Source (Reported vs. Reproduced)=Reported2026.0270.6
Claude-3.5-SonnetParadigm=LLM-as-a-Judge & Generative Reward Models2025.0770.6
Claude-3-5-sonnet-20240620Reward Model Category=GenRMs2025.0570.6
INF-ORM-Llama3.1-70BParadigm=Bradley-Terry2025.0770.5
INF-ORM-Llama3.1-70B2025.0770.5
INF-ORM-Llama3.1-70BReward Model Category=ScalarRMs2025.0570.5
DeepSeek-GRM-27BParadigm=LLM-as-a-Judge & Generative Reward Models, Integration=w/ MetaRM2025.0770.3
DeepSeek-GRM-27BGT=True2026.0470.3
FsfairX-LLaMA3-RM-v0.1Paradigm=Bradley-Terry2025.0770.2
GRM-Llama3-8B-rewardmodel-ftParadigm=Bradley-Terry2025.0770.2
CIR-AMS/BTRM_Qwen2_7b_06132025.0770.2
OPRM-RgFT-Qwen2.5-7BModel=OPRM, Backbone=Qwen2.5-7B, Region Flooding Tuning (RgFT)=true, Source (Reported vs. Reproduced)=Ours2026.0270.1
Nemotron-4-340B-RewardModel=Nemotron-4-340B-Reward, Source (Reported vs. Reproduced)=Reported2026.0269.9
RM-R1-7B (Instruct)Stage=SFT, RL, Data=9K, 64K2026.0369.9
Nemotron-4-340B-RewardReward Model Category=ScalarRMs2025.0569.9
RM-R1-DeepSeek-Distill-Qwen-32BParadigm=LLM-as-a-Judge & Generative Reward Models2025.0769.8
RM-R1-DEEPSEEK-DISTILLED-QWEN-32BReward Model Category=Our Methods2025.0569.8
Skywork-Reward-Gemma-2-27BParadigm=Bradley-Terry2025.0769.5
Skywork-Reward-Gemma-2-27B-v0.2Paradigm=Bradley-Terry2025.0769.4
Skywork-Reward-Llama-3.1-8BReward Model Category=ScalarRMs2025.0569.3
Skywork-Reward-Llama-3.1-8BParadigm=Bradley-Terry2025.0769.2
RM-R1-QWEN-INSTRUCT-14BReward Model Category=Our Methods2025.0569.2
DeepSeek-GRM-27BModel=DeepSeek-GRM-27B, Source (Reported vs. Reproduced)=Reproduced2026.0269
DeepSeek-GRM-27BParadigm=LLM-as-a-Judge & Generative Reward Models2025.0769
DeepSeek-GRM-27BReward Model Category=ReasRMs2025.0569
LLaMA-3.1-70b-InstructModel=LLaMA-3.1-70b-Instruct, Source (Reported vs. Reproduced)=Reported2026.0268.9
Llama3.1-70B-InstructReward Model Category=GenRMs2025.0568.9
GRM-llama3-8B-distillParadigm=Bradley-Terry2025.0768.8
RM-R1-DEEPSEEK-DISTILLED-QWEN-14BReward Model Category=Our Methods2025.0568.5
Eurus-RM-7bReward Model Category=ScalarRMs2025.0568.3
GRM-gemma2-2B-rewardmodel-ftParadigm=Bradley-Terry2025.0768
LDL-Reward-Gemma-2-27B-v0.1Paradigm=Bradley-Terry2025.0767.9
Internlm2-7b-rewardParadigm=Bradley-Terry2025.0767.1
Internlm2-7b-rewardReward Model Category=ScalarRMs2025.0567.1
DeepSeek-GRM-27B-RFTModel=DeepSeek-GRM-27B-RFT, Source (Reported vs. Reproduced)=Reproduced2026.0267
DeepSeek-GRM-27B-RFTReward Model Category=ReasRMs2025.0567
Self-taught-evaluator-llama3.1-70BReward Model Category=ReasRMs2025.0567
RM-Mistral-7BParadigm=Bradley-Terry2025.0766.6
Skywork-Reward-Llama-3.1-8B-v0.2Paradigm=Bradley-Terry2025.0766.6
RM-R1-QWEN-INSTRUCT-7BReward Model Category=Our Methods2025.0566.4
DeepSeek-V2.5-0905Model=DeepSeek-V2.5-0905, Source (Reported vs. Reproduced)=Reported2026.0265.7
URM-LLaMa-3.1-8BParadigm=Bradley-Terry2025.0765.7
Eurus-RM-7bParadigm=Bradley-Terry2025.0765.5
Ray2333/GRM-gemma2-2B-rewardmodel-ft2025.0765.5
Skywork-Critic-Llama-3.1-70BReward Model Category=GenRMs2025.0565.5
RM-R1-7B (Distill)Stage=SFT, RL, Data=9K, 64K2026.0365.1
Llama-3.1-Nemotron-70BParadigm=Bradley-Terry2025.0764.9
LLM-as-a-JudgeModel=LLM-as-a-Judge, Source (Reported vs. Reproduced)=Reproduced2026.0264.8
QRM-Llama3.1-8BParadigm=Bradley-Terry2025.0764.7
ArmoRM-8B-v0.1Model=ArmoRM-8B-v0.1, Source (Reported vs. Reproduced)=Reported2026.0264.6
ArmoRM-Llama3-8B-v0.1Paradigm=Bradley-Terry2025.0764.6