List your company
ResearchFormerly WizWand

Reward Modeling on PPE-P (Acc, Brier, CE)

63AccuracyTwo-anchor-Skywork
56.03257.84159.6561.459May 12, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Two-anchor-SkyworkBackbone=Llama-3.2-3B-Instruct, Training Dataset=HelpSteer3-20K2026.05630.2440.782
Two-anchor-DeepSeek-ProBackbone=Llama-3.2-8B-Instruct, Training Dataset=HelpSteer2-9K2026.0561.80.23350.6757
Two-anchor-SkyworkBackbone=Llama-3.2-8B-Instruct, Training Dataset=HelpSteer2-9K2026.0561.50.23030.6586
Two-anchor-DeepSeek-ProBackbone=Llama-3.2-3B-Instruct, Training Dataset=HelpSteer3-20K2026.0561.40.2440.732
Two-anchor-DeepSeek-FlashBackbone=Llama-3.2-3B-Instruct, Training Dataset=HelpSteer3-20K2026.0561.40.2460.742
Two-anchor-SkyworkBackbone=Llama-3.2-8B-Instruct, Training Dataset=MultiPref-8K2026.0561.30.22840.6523
Two-anchor-DeepSeek-FlashBackbone=Llama-3.2-8B-Instruct, Training Dataset=HelpSteer2-9K2026.0561.30.22920.6508
Two-anchor-SkyworkBackbone=Llama-3.2-3B-Instruct, Training Dataset=PersonalLLM-40K2026.0561.20.240.696
Two-anchor-DeepSeek-FlashBackbone=Llama-3.2-8B-Instruct, Training Dataset=MultiPref-8K2026.05610.23010.6538
GaussianBackbone=Llama-3.2-8B-Instruct, Training Dataset=MultiPref-8K2026.0560.90.23260.66
BTBackbone=Llama-3.2-3B-Instruct, Training Dataset=HelpSteer3-20K2026.0560.90.2670.862
Two-anchor-DeepSeek-FlashBackbone=Llama-3.2-3B-Instruct, Training Dataset=PersonalLLM-40K2026.0560.80.2380.68
Two-anchor-DeepSeek-ProBackbone=Llama-3.2-3B-Instruct, Training Dataset=PersonalLLM-40K2026.0560.70.2410.688
Two-anchor-DeepSeek-ProBackbone=Llama-3.2-8B-Instruct, Training Dataset=MultiPref-8K2026.0560.50.23210.6581
BTBackbone=Llama-3.2-8B-Instruct, Training Dataset=MultiPref-8K2026.0560.40.23410.663
BTBackbone=Llama-3.2-8B-Instruct, Training Dataset=HelpSteer2-9K2026.0560.40.2480.7315
GaussianBackbone=Llama-3.2-8B-Instruct, Training Dataset=HelpSteer2-9K2026.0560.40.24720.754
BT(Hard)Backbone=Llama-3.2-3B-Instruct, Training Dataset=HelpSteer3-20K2026.0560.30.2851.015
GaussianBackbone=Llama-3.2-3B-Instruct, Training Dataset=HelpSteer3-20K2026.0560.30.2640.932
Two-anchor-SkyworkBackbone=Llama-3.2-3B-Instruct, Training Dataset=HelpSteer2-9K2026.0560.20.2410.694
BT(Hard)Backbone=Llama-3.2-8B-Instruct, Training Dataset=HelpSteer2-9K2026.05600.23510.6688
Two-anchor-SkyworkBackbone=Llama-3.2-3B-Instruct, Training Dataset=MultiPref-8K2026.05600.2340.664
Two-anchor-DeepSeek-FlashBackbone=Llama-3.2-3B-Instruct, Training Dataset=HelpSteer2-9K2026.05600.2370.672
BT(Hard)Backbone=Llama-3.2-8B-Instruct, Training Dataset=MultiPref-8K2026.0559.90.27480.911
BT(Hard)Backbone=Llama-3.2-3B-Instruct, Training Dataset=PersonalLLM-40K2026.0559.50.3051.251
GaussianBackbone=Llama-3.2-3B-Instruct, Training Dataset=PersonalLLM-40K2026.0559.50.2530.725
Two-anchor-DeepSeek-ProBackbone=Llama-3.2-3B-Instruct, Training Dataset=HelpSteer2-9K2026.0559.40.2340.66
BTBackbone=Llama-3.2-3B-Instruct, Training Dataset=MultiPref-8K2026.0559.30.2360.665
BTBackbone=Llama-3.2-3B-Instruct, Training Dataset=PersonalLLM-40K2026.0559.20.2520.729
BTBackbone=Llama-3.2-3B-Instruct, Training Dataset=HelpSteer2-9K2026.0558.80.2510.719
BT(Hard)Backbone=Llama-3.2-3B-Instruct, Training Dataset=HelpSteer2-9K2026.0558.40.2840.932
Two-anchor-DeepSeek-FlashBackbone=Llama-3.2-3B-Instruct, Training Dataset=MultiPref-8K2026.0558.20.2390.672
GaussianBackbone=Llama-3.2-3B-Instruct, Training Dataset=HelpSteer2-9K2026.0558.20.2510.729
GaussianBackbone=Llama-3.2-3B-Instruct, Training Dataset=MultiPref-8K2026.0557.70.2420.68
BT(Hard)Backbone=Llama-3.2-3B-Instruct, Training Dataset=MultiPref-8K2026.0557.10.2950.962
Two-anchor-DeepSeek-ProBackbone=Llama-3.2-3B-Instruct, Training Dataset=MultiPref-8K2026.0556.30.2440.682