List your company

Over-refusal Rate Analysis on OR-Bench

16.68Over-refusal RateAdaCD
13.374435.68725880.3128Jan 15, 2026Jan 30, 2026Feb 15, 2026Mar 2, 2026Mar 18, 2026Apr 2, 2026Apr 18, 2026

Evaluation results

MethodLinks
AdaCDBase Model=Qwen32026.0416.68
SelfCDBase Model=Qwen32026.0418.2
SurgicalBase Model=Qwen32026.0422.52
SSPModel Backbone=Mistral3-8B2026.0124.1
SSPModel Backbone=LLaMA3-8B2026.0124.6
SSPModel Backbone=Qwen2.5-7B2026.0125.3
ACE-SafetyModel Backbone=Mistral3-8B2026.0128
SafeDecodingModel Backbone=Mistral3-8B2026.0128.7
ACE-SafetyModel Backbone=LLaMA3-8B2026.0128.7
SafeDecodingModel Backbone=LLaMA3-8B2026.0129.3
ACE-SafetyModel Backbone=Qwen2.5-7B2026.0129.6
SafeDecodingModel Backbone=Qwen2.5-7B2026.0130.1
DefaultBase Model=Qwen32026.0430.63
AdaCDBase Model=Llama32026.0430.93
SmoothLLMModel Backbone=Mistral3-8B2026.0132.6
SmoothLLMModel Backbone=LLaMA3-8B2026.0133.2
Self-ReminderModel Backbone=Mistral3-8B2026.0134.5
SmoothLLMModel Backbone=Qwen2.5-7B2026.0134.5
SSDBase Model=Qwen32026.0435.03
Self-ReminderModel Backbone=LLaMA3-8B2026.0135.1
Self-ReminderModel Backbone=Qwen2.5-7B2026.0136.2
SelfCDBase Model=Llama32026.0437.15
SurgicalBase Model=Llama32026.0441.32
PromptBase Model=Qwen32026.0453.75
SurgicalBase Model=Gemma22026.0460.5
AdaCDBase Model=Gemma22026.0462.55
DefaultBase Model=Llama32026.0467.17
SelfCDBase Model=Gemma22026.0469.14
PromptBase Model=Gemma22026.0474.6
SSDBase Model=Gemma22026.0475.97
DefaultBase Model=Gemma22026.0478.7
PromptBase Model=Llama32026.0492.95
SSDBase Model=Llama32026.0499.32