List your company

General Knowledge on MMLU (Drop in Utility %)

37.4MMLU General Knowledge Drop in Utility (%)KL
-37.376-17.9631.4520.863Apr 1, 2026

Evaluation results

MethodLinks
KLAlignment Strategy=Deliberative Alignment (GRPO)2026.0437.4
Latent Similarity- Layer 12Alignment Strategy=Deliberative Alignment (GRPO)2026.049.7
Self CertaintyAlignment Strategy=Deliberative Alignment (GRPO)2026.041.3
RandomAlignment Strategy=Deliberative Alignment (GRPO)2026.040.7
PerplexityAlignment Strategy=Deliberative Alignment (GRPO)2026.04-6.7
Latent Similarity- Best LayerAlignment Strategy=Deliberative Alignment (GRPO)2026.04-34.5