Any other benchmarks like MMLU or other coding benches
#2
by
NoelJacob
- opened
Same
This is a fine-tune for a specific reasoning task. I don't think it would do better than the base on MMLU or any coding benches.
Yes. I understand, but there's a chance it might. Grok 3 was mostly only trained on reasoning of coding but it made it perform well on general reasoning.