List your company

Over-refusal Evaluation on ORFuzzSet

16ORRLLM-VA
12.6435.325880.68Jan 27, 2026

Evaluation results

MethodLinks
LLM-VAModel=Qwen2.5, Size=3B2026.0116
LLM-VAModel=Qwen2.5, Size=7B2026.0122
LLM-VAModel=Qwen3, Size=4B2026.0124
LLM-VAModel=Qwen3, Size=8B2026.0124
SCANSModel=Mistral-v0.3, Size=7B2026.0128
LLM-VAModel=Mistral-v0.3, Size=7B2026.0128
LLM-VAModel=gemma-2, Size=9B2026.0136
LLM-VAModel=Llama-3.1, Size=8B2026.0138
LLM-VAModel=Phi-4, Size=15B2026.0138
VectorSteerModel=Qwen3, Size=8B2026.0140
VectorSteerModel=Mistral-v0.3, Size=7B2026.0144
AlphaSteerModel=Mistral-v0.3, Size=7B2026.0146
AlphaSteerModel=Qwen3, Size=14B2026.0146
LLM-VAModel=Phi-4, Size=4B2026.0148
AlphaSteer+Model=Qwen3, Size=4B2026.0148
VectorSteerModel=Qwen3, Size=4B2026.0148
LLM-VAModel=Phi-3.5, Size=4B2026.0150
AlphaSteer+Model=Mistral-v0.3, Size=7B2026.0152
AlphaSteer+Model=Qwen3, Size=14B2026.0152
OriginalModel=Mistral-v0.3, Size=7B2026.0154
AlphaSteerModel=Qwen3, Size=4B2026.0156
LLM-VAModel=Qwen3, Size=14B2026.0156
AlphaSteer+Model=Qwen2.5, Size=3B2026.0158
AlphaSteerModel=Qwen3, Size=8B2026.0158
AlphaSteer+Model=Qwen3, Size=8B2026.0160
OriginalModel=Qwen2.5, Size=3B2026.0162
AlphaSteerModel=Qwen2.5, Size=3B2026.0162
LLM-VAModel=Qwen2.5, Size=14B2026.0166
OriginalModel=Qwen3, Size=4B2026.0166
SCANSModel=Qwen3, Size=4B2026.0166
OriginalModel=Phi-4, Size=4B2026.0168
VectorSteerModel=Phi-4, Size=4B2026.0168
AlphaSteerModel=Phi-4, Size=4B2026.0170
SCANSModel=Qwen2.5, Size=3B2026.0170
OriginalModel=Qwen3, Size=8B2026.0172
VectorSteerModel=Qwen3, Size=14B2026.0172
AlphaSteer+Model=Phi-4, Size=4B2026.0174
VectorSteerModel=Phi-3.5, Size=4B2026.0178
SCANSModel=Phi-4, Size=4B2026.0178
OriginalModel=Qwen2.5, Size=7B2026.0180
VectorSteerModel=Qwen2.5, Size=7B2026.0180
AlphaSteer+Model=Qwen2.5, Size=7B2026.0182
SCANSModel=Qwen2.5, Size=14B2026.0182
SCANSModel=Qwen3, Size=14B2026.0182
SCANSModel=Qwen3, Size=8B2026.0184
AlphaSteerModel=Phi-3.5, Size=4B2026.0186
AlphaSteerModel=Qwen2.5, Size=7B2026.0186
SCANSModel=Qwen2.5, Size=7B2026.0186
OriginalModel=Qwen3, Size=14B2026.0186
AlphaSteer+Model=Phi-3.5, Size=4B2026.0188
OriginalModel=Phi-3.5, Size=4B2026.0190
OriginalModel=Qwen2.5, Size=14B2026.0190
SCANSModel=gemma-2, Size=9B2026.0192
VectorSteerModel=Qwen2.5, Size=3B2026.0192
AlphaSteer+Model=Qwen2.5, Size=14B2026.0192
AlphaSteerModel=Qwen2.5, Size=14B2026.0192
AlphaSteer+Model=gemma-2, Size=9B2026.0194
VectorSteerModel=Phi-4, Size=15B2026.0194
SCANSModel=Phi-4, Size=15B2026.0194
SCANSModel=Phi-3.5, Size=4B2026.0196
AlphaSteer+Model=Phi-4, Size=15B2026.0196
AlphaSteerModel=Phi-4, Size=15B2026.0196
VectorSteerModel=Qwen2.5, Size=14B2026.0196
OriginalModel=gemma-2, Size=9B2026.0198
AlphaSteerModel=gemma-2, Size=9B2026.0198
VectorSteerModel=gemma-2, Size=9B2026.0198
OriginalModel=Phi-4, Size=15B2026.0198
OriginalModel=Llama-3.1, Size=8B2026.01100
AlphaSteer+Model=Llama-3.1, Size=8B2026.01100
AlphaSteerModel=Llama-3.1, Size=8B2026.01100
VectorSteerModel=Llama-3.1, Size=8B2026.01100
SCANSModel=Llama-3.1, Size=8B2026.01100