List your company

Commonsense Reasoning on HellaSwag (Accuracy, Delta)

0.061DeltaKICK-SETTLE
-0.0023360.0141070.030550.046993Jul 12, 2024Oct 31, 2024Feb 20, 2025Jun 12, 2025Oct 2, 2025Jan 22, 2026May 14, 2026

Evaluation results

MethodLinks
KICK-SETTLEDefense=SDD, Model=Llama-3.1-8B, Attack Type=Kick-settle2026.050.061—
KICK-SETTLEDefense=SDD, Model=Llama-2-7B, Attack Type=Kick-settle2026.050.041—
KICK-SETTLEDefense=SDD, Model=Qwen3-8B, Attack Type=Kick-settle2026.050.005—
FlashNormModel=Llama-3.1-8B, Few-shot setting=0-shot, Context window=2048-token, Precision=fp16, Hardware=NVIDIA A100, Evaluation framework=lm-evaluation-harness2024.070.000160.72
SourceModel=Llama-3.1-8B, Few-shot setting=0-shot, Context window=2048-token, Precision=fp16, Hardware=NVIDIA A100, Evaluation framework=lm-evaluation-harness2024.07—60.71