List your company
ResearchFormerly WizWand

Reward Modeling on PPE-P latest (test)

60.2AccuracyTwo-anchor-Skywork
55.41656.65857.959.142May 12, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Two-anchor-SkyworkBackbone=Llama-3.2-1B-Instruct, Training Dataset=HelpSteer3-20K2026.0560.20.25350.7878
BTBackbone=Llama-3.2-1B-Instruct, Training Dataset=HelpSteer3-20K2026.05590.2760.8827
BT(Hard)Backbone=Llama-3.2-1B-Instruct, Training Dataset=HelpSteer3-20K2026.05590.29081.0017
Two-anchor-DeepSeek-FlashBackbone=Llama-3.2-1B-Instruct, Training Dataset=HelpSteer3-20K2026.05590.25520.748
Two-anchor-SkyworkBackbone=Llama-3.2-1B-Instruct, Training Dataset=PersonalLLM-40K2026.0558.80.24530.702
BT(Hard)Backbone=Llama-3.2-1B-Instruct, Training Dataset=PersonalLLM-40K2026.0558.50.30971.2321
Two-anchor-DeepSeek-FlashBackbone=Llama-3.2-1B-Instruct, Training Dataset=PersonalLLM-40K2026.0558.40.24830.7065
Two-anchor-DeepSeek-ProBackbone=Llama-3.2-1B-Instruct, Training Dataset=HelpSteer3-20K2026.0558.30.25780.7609
Two-anchor-DeepSeek-ProBackbone=Llama-3.2-1B-Instruct, Training Dataset=PersonalLLM-40K2026.0558.20.24910.7064
GaussianBackbone=Llama-3.2-1B-Instruct, Training Dataset=HelpSteer3-20K2026.05580.27450.9378
GaussianBackbone=Llama-3.2-1B-Instruct, Training Dataset=PersonalLLM-40K2026.0557.80.25950.7519
BTBackbone=Llama-3.2-1B-Instruct, Training Dataset=PersonalLLM-40K2026.0557.50.26160.7592
BT(Hard)Backbone=Llama-3.2-1B-Instruct, Training Dataset=HelpSteer2-9K2026.0557.40.24140.6783
Two-anchor-SkyworkBackbone=Llama-3.2-1B-Instruct, Training Dataset=HelpSteer2-9K2026.0557.40.24070.6761
BTBackbone=Llama-3.2-1B-Instruct, Training Dataset=HelpSteer2-9K2026.0557.30.24130.6786
Two-anchor-DeepSeek-FlashBackbone=Llama-3.2-1B-Instruct, Training Dataset=HelpSteer2-9K2026.0556.90.25220.7135
GaussianBackbone=Llama-3.2-1B-Instruct, Training Dataset=MultiPref-8K2026.0556.80.24580.6873
Two-anchor-SkyworkBackbone=Llama-3.2-1B-Instruct, Training Dataset=MultiPref-8K2026.0556.80.24650.6925
Two-anchor-DeepSeek-ProBackbone=Llama-3.2-1B-Instruct, Training Dataset=HelpSteer2-9K2026.0556.70.24130.6766
GaussianBackbone=Llama-3.2-1B-Instruct, Training Dataset=HelpSteer2-9K2026.0556.30.2630.768
BT(Hard)Backbone=Llama-3.2-1B-Instruct, Training Dataset=MultiPref-8K2026.0556.20.28760.8883
Two-anchor-DeepSeek-ProBackbone=Llama-3.2-1B-Instruct, Training Dataset=MultiPref-8K2026.0556.20.24280.6793
Two-anchor-DeepSeek-FlashBackbone=Llama-3.2-1B-Instruct, Training Dataset=MultiPref-8K2026.0555.70.24410.6828
BTBackbone=Llama-3.2-1B-Instruct, Training Dataset=MultiPref-8K2026.0555.60.24810.6958