List your company

Mathematical Reasoning on MMLU Out of Domain

1,896.9MTL ScoreFARL
1,179.7161,365.9081,552.11,738.292Sep 29, 2025

Evaluation results

MethodLinks
FARLBackbone=R1-Llama-8B, Training Time=4h 26m 4s2025.091,896.975.7
RL (GRPO)Backbone=R1-Llama-8B, Training Time=4h 6m 27s2025.091,844.474.5
R1-Llama-8B (Base)Backbone=R1-Llama-8B2025.091,386.271.6
SFTBackbone=R1-Llama-8B, Training Time=10m 21s2025.091,207.373.2