List your company
ResearchFormerly WizWand

Relative Robustness Analysis on Combined Past Tense, OR-Bench, MMLU

78.9R-ScoreOnly Scaling
3.000822.705442.4162.1146Sep 30, 2025
Updated 6mo ago

Evaluation results

MethodLinks
Only ScalingBase Model=Qwen2.5-7B-Instruct2025.0978.946.4
SFT (5/95)Base Model=Qwen2.5-7B-Instruct2025.0975.639.8
ASGUARDBase Model=Qwen2.5-7B-Instruct2025.0974.658.8
ASGUARDBase Model=OLMo-2-1124-7B-Instruct2025.0973.746.3
RepBendBase Model=Gemma-2-9B-it2025.0972.841.9
SFT (5/95)Base Model=Llama-3.1-8B-Instruct2025.0972.446.7
ASGUARDBase Model=Llama-3.1-8B-Instruct2025.0971.852.9
Only ScalingBase Model=Llama-3.1-8B-Instruct2025.0971.650.3
RepEBase Model=Gemma-2-9B-it2025.0970.537.2
ASGUARDBase Model=Gemma-2-9B-it2025.0970.144.6
DPOBase Model=Gemma-2-9B-it2025.0969.835.4
DPOBase Model=Llama-3.1-8B-Instruct2025.0969.536.7
SFT (30/70)Base Model=OLMo-2-1124-7B-Instruct2025.0968.644.3
CBBase Model=Gemma-2-9B-it2025.096835
Head AblationBase Model=Gemma-2-9B-it2025.0967.936
SFT (5/95)Base Model=OLMo-2-1124-7B-Instruct2025.0967.637.3
CBBase Model=Qwen2.5-7B-Instruct2025.0967.135.5
Head AblationBase Model=Qwen2.5-7B-Instruct2025.0966.938.5
DPOBase Model=OLMo-2-1124-7B-Instruct2025.0966.935
SFT (30/70)Base Model=Qwen2.5-7B-Instruct2025.0966.458.7
RepEBase Model=Qwen2.5-7B-Instruct2025.0966.335.7
CBBase Model=OLMo-2-1124-7B-Instruct2025.096637
RepBendBase Model=Llama-3.1-8B-Instruct2025.0965.748.4
DPOBase Model=Qwen2.5-7B-Instruct2025.0965.533.8
Head AblationBase Model=OLMo-2-1124-7B-Instruct2025.0965.535.8
RepBendBase Model=OLMo-2-1124-7B-Instruct2025.0964.934.9
RepEBase Model=Llama-3.1-8B-Instruct2025.0964.532.8
Only ScalingBase Model=OLMo-2-1124-7B-Instruct2025.0964.337.7
RepEBase Model=OLMo-2-1124-7B-Instruct2025.0961.933.9
RepBendBase Model=Qwen2.5-7B-Instruct2025.0960.240.6
SFT (5/95)Base Model=Gemma-2-9B-it2025.0958.648.3
Head AblationBase Model=Llama-3.1-8B-Instruct2025.0957.335.1
SFT (30/70)Base Model=Gemma-2-9B-it2025.095647
SFT (30/70)Base Model=Llama-3.1-8B-Instruct2025.0952.245.6
CBBase Model=Llama-3.1-8B-Instruct2025.0930.636.3
Only ScalingBase Model=Gemma-2-9B-it2025.095.928.96