Per-subject MMLU scores
#4
by
onuralp
- opened
Congrats on this powerful model. Any chance you can share the per-subject MMLU scores? I'd like to better understand the correlation between TruthfulQA performance and per-subject MMLU accuracy.
Congrats on this powerful model. Any chance you can share the per-subject MMLU scores? I'd like to better understand the correlation between TruthfulQA performance and per-subject MMLU accuracy.
There is in the HF Leaderboard repo files somewhere
Sweet! I did not know that the detailed logs were available. Thanks!
For future reference, here is the detailed log for this model.
onuralp
changed discussion status to
closed