diff options
| author | YurenHao0426 <blackhao0426@gmail.com> | 2026-06-13 12:35:36 -0500 |
|---|---|---|
| committer | YurenHao0426 <blackhao0426@gmail.com> | 2026-06-13 12:35:36 -0500 |
| commit | 66e0d8b9fd4d0f7a2231d689c055e26fdf1cf04a (patch) | |
| tree | c29cba61124018755a19b02c9d33e3ad5f2e05cc /research/flossing/multi4_eval_compare/trm_multi4_eval.csv | |
Curated export for clone-and-run Maze training (2x A6000) + diagnostics.
trm/hrm pretrain.py carry trajectory-augmentation code (backward-compatible).
Heavy artifacts (checkpoints/wandb/npz) gitignored; see PROVENANCE.md.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Diffstat (limited to 'research/flossing/multi4_eval_compare/trm_multi4_eval.csv')
| -rw-r--r-- | research/flossing/multi4_eval_compare/trm_multi4_eval.csv | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/research/flossing/multi4_eval_compare/trm_multi4_eval.csv b/research/flossing/multi4_eval_compare/trm_multi4_eval.csv new file mode 100644 index 0000000..f3d16b5 --- /dev/null +++ b/research/flossing/multi4_eval_compare/trm_multi4_eval.csv @@ -0,0 +1,2 @@ +run,step,all/exact_accuracy,all/accuracy,all/lm_loss,all/q_halt_accuracy,all/q_halt_loss,all/steps
+trm_multi4,26041,0.7394047975540161,0.9067130088806152,0.21417337656021118,0.9997232556343079,0.0024572687689214945,16
|