List your company

Safety-Utility Trade-off Evaluation on S-Eval, ORFuzzSet, and NQ Aggregated

86.81F1 ScoreLLM-VA
46.19856.741567.28577.8285Jan 27, 2026

Evaluation results

MethodLinks
LLM-VAModel=gemma-2, Size=9B2026.0186.81
LLM-VAModel=Phi-4, Size=15B2026.0185.26
LLM-VAModel=Qwen3, Size=8B2026.0183.81
LLM-VAModel=Llama-3.1, Size=8B2026.0181.72
VectorSteerModel=Qwen3, Size=8B2026.0180.2
VectorSteerModel=Qwen3, Size=4B2026.0179.79
LLM-VAModel=Qwen2.5, Size=3B2026.0179.25
AlphaSteerModel=Qwen3, Size=14B2026.0178.97
LLM-VAModel=Qwen3, Size=4B2026.0178.22
SCANSModel=Mistral-v0.3, Size=7B2026.0177.42
LLM-VAModel=Mistral-v0.3, Size=7B2026.0176.56
LLM-VAModel=Qwen2.5, Size=7B2026.0175.98
VectorSteerModel=Mistral-v0.3, Size=7B2026.0175
AlphaSteerModel=Qwen3, Size=8B2026.0174.74
VectorSteerModel=Qwen3, Size=14B2026.0173.99
AlphaSteer+Model=Qwen3, Size=4B2026.0173.53
AlphaSteer+Model=Qwen2.5, Size=3B2026.0173.33
AlphaSteerModel=Mistral-v0.3, Size=7B2026.0172.73
AlphaSteer+Model=Qwen3, Size=14B2026.0172.55
SCANSModel=Qwen3, Size=4B2026.0171.35
AlphaSteerModel=Qwen3, Size=4B2026.0171.29
LLM-VAModel=Qwen3, Size=14B2026.0171.29
AlphaSteerModel=Phi-4, Size=4B2026.0171.26
AlphaSteer+Model=Mistral-v0.3, Size=7B2026.0171.22
VectorSteerModel=Phi-4, Size=4B2026.0171.19
AlphaSteer+Model=Qwen3, Size=8B2026.0171.04
AlphaSteerModel=Qwen2.5, Size=3B2026.0170.72
AlphaSteer+Model=Phi-4, Size=4B2026.0169.77
LLM-VAModel=Qwen2.5, Size=14B2026.0169.11
SCANSModel=Qwen2.5, Size=3B2026.0168.89
LLM-VAModel=Phi-3.5, Size=4B2026.0168.22
AlphaSteerModel=Qwen2.5, Size=14B2026.0167.95
AlphaSteerModel=Phi-4, Size=15B2026.0167.11
VectorSteerModel=Qwen2.5, Size=14B2026.0167.1
VectorSteerModel=Phi-4, Size=15B2026.0166.67
AlphaSteer+Model=Phi-4, Size=15B2026.0166.23
AlphaSteer+Model=Qwen2.5, Size=7B2026.0165.54
LLM-VAModel=Phi-4, Size=4B2026.0165.45
AlphaSteerModel=Qwen2.5, Size=7B2026.0164.37
AlphaSteer+Model=gemma-2, Size=9B2026.0164.15
SCANSModel=Phi-4, Size=15B2026.0164.1
AlphaSteer+Model=Qwen2.5, Size=14B2026.0163.86
VectorSteerModel=Phi-3.5, Size=4B2026.0163.8
AlphaSteerModel=Llama-3.1, Size=8B2026.0163.51
SCANSModel=Qwen3, Size=8B2026.0163.1
AlphaSteer+Model=Phi-3.5, Size=4B2026.0162.5
AlphaSteerModel=gemma-2, Size=9B2026.0162.42
SCANSModel=Qwen2.5, Size=14B2026.0162.15
AlphaSteerModel=Phi-3.5, Size=4B2026.0161.9
OriginalModel=Phi-4, Size=15B2026.0161.82
SCANSModel=Qwen2.5, Size=7B2026.0161.45
AlphaSteer+Model=Llama-3.1, Size=8B2026.0161.22
VectorSteerModel=Llama-3.1, Size=8B2026.0161.11
OriginalModel=Llama-3.1, Size=8B2026.0161.04
OriginalModel=Qwen3, Size=4B2026.0159.56
SCANSModel=Llama-3.1, Size=8B2026.0159.31
OriginalModel=Phi-4, Size=4B2026.0159.18
SCANSModel=gemma-2, Size=9B2026.0158.9
VectorSteerModel=Qwen2.5, Size=7B2026.0158.54
OriginalModel=Qwen3, Size=8B2026.0157.53
OriginalModel=Qwen2.5, Size=3B2026.0157.41
VectorSteerModel=gemma-2, Size=9B2026.0157.14
OriginalModel=Qwen2.5, Size=14B2026.0156.68
OriginalModel=Mistral-v0.3, Size=7B2026.0156.35
SCANSModel=Phi-4, Size=4B2026.0155.13
OriginalModel=Qwen3, Size=14B2026.0153.02
OriginalModel=Qwen2.5, Size=7B2026.0152.97
OriginalModel=Phi-3.5, Size=4B2026.0150.73
VectorSteerModel=Qwen2.5, Size=3B2026.0150.67
OriginalModel=gemma-2, Size=9B2026.0149.14
SCANSModel=Qwen3, Size=14B2026.0147.85
SCANSModel=Phi-3.5, Size=4B2026.0147.76