List your company

Pretraining Data Mixture Estimation on OLMo Pretraining Mixture (Temporal held-out)

95.5Web Source EstimateLLMSurgeon
76.135281.162686.1991.2174May 28, 2026

Evaluation results

MethodLinks
LLMSurgeonModel=OLMo-22026.0595.52.1101.50.50.389.18
LLMSurgeonModel=OLMo-1B2026.0581.113.4100.22.32.994.46
LLMSurgeonModel=OLMo-3 (held-out)2026.0576.887.06412.60.822.686.41