List your company

Safety and Helpfulness Evaluation on StrongREJECT

0.2Harm RatePPO SafeSearch
-2.60816.34635.354.254Oct 19, 2025

Evaluation results

MethodLinks
PPO SafeSearchBackbone=Qwen-2.5-7B-Instruct2025.100.22.65
PPO SafeSearch w/o qr. & hf.Backbone=Qwen-2.5-7B-Instruct2025.100.21.61
GRPO SafeSearchBackbone=Qwen-2.5-7B-Instruct2025.100.52.59
PPO SafeSearch w/o hf.Backbone=Qwen-2.5-7B-Instruct2025.101.61.7
Base LLMBackbone=Mistral-NeMo-8B-Instruct2025.101.93.32
PPO SafeSearch w/o qr.Backbone=Qwen-2.5-7B-Instruct2025.105.82.36
Base AgentBackbone=Qwen-2.5-14B-Instruct2025.107.91.84
Base LLMBackbone=Qwen-2.5-14B-Instruct2025.108.81.78
Base AgentBackbone=Mistral-NeMo-8B-Instruct2025.1011.23.15
Naive RAGBackbone=Mistral-NeMo-8B-Instruct2025.1011.43.26
Base LLMBackbone=Qwen-2.5-7B-Instruct2025.1014.32.15
PPO SafeSearchBackbone=Qwen-2.5-3B-Instruct2025.10152.22
Naive RAGBackbone=Qwen-2.5-14B-Instruct2025.1017.31.74
GRPO Utility-Only Ft.Backbone=Qwen-2.5-7B-Instruct2025.10201.96
BaseBackbone=Qwen-2.5-7B-Instruct2025.1027.32.19
Base AgentBackbone=Qwen-2.5-7B-Instruct2025.1027.32.19
Search BlockBackbone=Qwen-2.5-7B-Instruct2025.1031.12.18
Base LLMBackbone=Qwen-2.5-3B-Instruct2025.1031.32.17
Document FilterBackbone=Qwen-2.5-7B-Instruct2025.1035.22.27
PPO Utility-Only Ft.Backbone=Qwen-2.5-7B-Instruct2025.1036.32.24
Ft. AgentBackbone=Qwen-2.5-7B-Instruct2025.1036.32.24
Naive RAGBackbone=Qwen-2.5-7B-Instruct2025.10512.17
BaseBackbone=Qwen-2.5-3B-Instruct2025.1052.72.33
Base AgentBackbone=Qwen-2.5-3B-Instruct2025.1052.72.33
Search BlockBackbone=Qwen-2.5-3B-Instruct2025.1055.32.2
Document FilterBackbone=Qwen-2.5-3B-Instruct2025.1062.12.21
PPO Utility-Only Ft.Backbone=Qwen-2.5-3B-Instruct2025.1063.52.22
Ft. AgentBackbone=Qwen-2.5-3B-Instruct2025.1063.52.22
Naive RAGBackbone=Qwen-2.5-3B-Instruct2025.1070.42.12