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/surrogate_flossing/figures/binary_surrogate_gradient_flossing.md | |
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/surrogate_flossing/figures/binary_surrogate_gradient_flossing.md')
| -rw-r--r-- | research/flossing/surrogate_flossing/figures/binary_surrogate_gradient_flossing.md | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/research/flossing/surrogate_flossing/figures/binary_surrogate_gradient_flossing.md b/research/flossing/surrogate_flossing/figures/binary_surrogate_gradient_flossing.md new file mode 100644 index 0000000..abc0c63 --- /dev/null +++ b/research/flossing/surrogate_flossing/figures/binary_surrogate_gradient_flossing.md @@ -0,0 +1,2 @@ + +**Figure:** Surrogate gradient flossing regularizes surrogate Lyapunov exponents and facilitates gradient signal propagation in binary neural networks. |