List your company

Safety and Helpfulness Evaluation on RRB

1HarmRGRPO SafeSearch
-1.01612.59226.239.808Oct 19, 2025

Evaluation results

MethodLinks
GRPO SafeSearchBackbone=Qwen-2.5-7B-Instruct2025.1012.7
PPO SafeSearchBackbone=Qwen-2.5-7B-Instruct2025.102.22.69
PPO SafeSearch w/o hf.Backbone=Qwen-2.5-7B-Instruct2025.102.61.81
PPO SafeSearch w/o qr. & hf.Backbone=Qwen-2.5-7B-Instruct2025.1031.84
PPO SafeSearch w/o qr.Backbone=Qwen-2.5-7B-Instruct2025.108.92.42
PPO SafeSearchBackbone=Qwen-2.5-3B-Instruct2025.1014.32.57
GRPO Utility-Only Ft.Backbone=Qwen-2.5-7B-Instruct2025.10232.29
BaseBackbone=Qwen-2.5-7B-Instruct2025.1029.12.52
Search BlockBackbone=Qwen-2.5-7B-Instruct2025.1030.52.49
Document FilterBackbone=Qwen-2.5-7B-Instruct2025.1031.72.53
PPO Utility-Only Ft.Backbone=Qwen-2.5-7B-Instruct2025.1032.42.55
BaseBackbone=Qwen-2.5-3B-Instruct2025.1039.82.6
Document FilterBackbone=Qwen-2.5-3B-Instruct2025.1049.52.6
Search BlockBackbone=Qwen-2.5-3B-Instruct2025.1050.52.54
PPO Utility-Only Ft.Backbone=Qwen-2.5-3B-Instruct2025.1051.42.58