List your company

Reward Modeling on RM-Bench Hard

0.7091Average AccuracyHARVE
0.43350.505050.57660.64815Jun 2, 2026

Evaluation results

MethodLinks
HARVEBackbone=Skywork-V2-Llama-3.2 (3B), alpha=1.252026.060.7091—0.64340.50580.62760.98940.8259
BaselineBackbone=Skywork-V2-Llama-3.2 (3B)2026.060.6943—0.55040.5190.62510.97540.7919
Finetune (3:1)Backbone=Skywork-V2-Llama-3.2 (3B), ratio=3:12026.060.694—0.5530.52190.62260.97540.7919
Finetune (5:1)Backbone=Skywork-V2-Llama-3.2 (3B), ratio=5:12026.060.6933—0.55040.5190.62260.97540.7919
HARVEBackbone=RM-Mistral (7B), alpha=2.502026.060.6506—0.53750.49270.52930.94250.8535
HARVEBackbone=InternLM2-reward (20B), alpha=1.602026.060.6378—0.3540.35530.66980.95890.5924
BaselineBackbone=InternLM2-reward (20B)2026.060.6265—0.2920.35820.67860.95660.518
Finetune (3:1)Backbone=InternLM2-reward (20B), ratio=3:12026.060.6257—0.2920.35960.67610.95660.518
Finetune (5:1)Backbone=InternLM2-reward (20B), ratio=5:12026.060.6247—0.28680.35670.67360.95770.5244
HARVEBackbone=Skywork-V2-Qwen3 (0.6B), alpha=1.252026.060.6204—0.49870.37130.57660.96010.6157
HARVEBackbone=Skywork-Reward-Llama-3.1 (8B), alpha=1.102026.060.6122—0.5090.45180.43040.98590.8662
BaselineBackbone=InternLM2-reward (1.8B)2026.060.5956—0.24290.25150.69690.92610.4459
HARVEBackbone=InternLM2-reward (1.8B), alpha=1.102026.060.5948—0.1990.37430.68240.90140.3907
Finetune (3:1)Backbone=InternLM2-reward (1.8B), ratio=3:12026.060.5933—0.24030.24560.69440.92610.4459
Finetune (5:1)Backbone=InternLM2-reward (1.8B), ratio=5:12026.060.5896—0.23510.23830.68750.92840.448
Finetune (3:1)Backbone=Skywork-V2-Qwen3 (0.6B), ratio=3:12026.060.5717—0.26870.43570.540.88850.552
Finetune (5:1)Backbone=Skywork-V2-Qwen3 (0.6B), ratio=5:12026.060.57—0.25320.43860.53940.8850.5541
BaselineBackbone=Skywork-V2-Qwen3 (0.6B)2026.060.569—0.26360.44150.54320.88970.5117
HARVEBackbone=GRM-Llama-3.2 (3B), alpha=1.252026.060.5659—0.64340.47080.28860.97890.828
Finetune (5:1)Backbone=Skywork-Reward-Llama-3.1 (8B), ratio=5:12026.060.5479—0.33590.3260.36670.98360.8662
Finetune (3:1)Backbone=Skywork-Reward-Llama-3.1 (8B), ratio=3:12026.060.5471—0.33330.3260.36610.98360.8641
BaselineBackbone=Skywork-Reward-Llama-3.1 (8B)2026.060.5468—0.32560.32750.3680.98360.8599
HARVEBackbone=FsfairX-LLaMA3-RM (8B), alpha=1.752026.060.5438—0.31270.43860.40580.96130.5966
Finetune (5:1)Backbone=GRM-Llama-3.2 (3B), ratio=5:12026.060.5318—0.32820.26460.36230.97420.8577
Finetune (3:1)Backbone=GRM-Llama-3.2 (3B), ratio=3:12026.060.5315—0.32820.26460.36170.97420.8577
BaselineBackbone=GRM-Llama-3.2 (3B)2026.060.5303—0.31780.26170.36230.9730.8599
Finetune (5:1)Backbone=FsfairX-LLaMA3-RM (8B), ratio=5:12026.060.5247—0.24550.39330.40770.95190.5669
Finetune (3:1)Backbone=FsfairX-LLaMA3-RM (8B), ratio=3:12026.060.5094—0.21710.37130.39630.92960.5711
BaselineBackbone=FsfairX-LLaMA3-RM (8B)2026.060.5046—0.20670.37130.39190.9190.5732
Finetune (5:1)Backbone=RM-Mistral (7B), ratio=5:12026.060.4788—0.24550.26610.33140.95190.62
Finetune (3:1)Backbone=RM-Mistral (7B), ratio=3:12026.060.4574—0.19120.26460.30750.93310.6008
BaselineBackbone=RM-Mistral (7B)2026.060.4441—0.16540.25290.30880.88620.6072
ArmoRM-Llama3-8B-v0.1Backbone=Llama3-8B2026.02—0.558—————
Athene-RM-8BSize=8B2026.02—0.514—————
INF-ORM-Llama3.1-70BBackbone=Llama3.1-70B2026.02—0.54—————
Internlm2-20b-rewardSize=20b2026.02—0.628—————
Llama-3-OffsetBias-RM-8BSize=8B2026.02—0.569—————
Llama-3.1-Nemotron-70BSize=70B2026.02—0.478—————
Skywork-Reward-Gemma-2-27B-v0.2Backbone=Gemma-2-27B2026.02—0.421—————
Skywork-Reward-Llama-3.1-8B-v0.2Backbone=Llama-3.1-8B2026.02—0.493—————
WILDREWARD-4BSize=4B2026.02—0.686—————
WILDREWARD-8BSize=8B2026.02—0.697—————