diff options
| author | YurenHao0426 <Blackhao0426@gmail.com> | 2026-03-25 10:23:19 -0500 |
|---|---|---|
| committer | YurenHao0426 <Blackhao0426@gmail.com> | 2026-03-25 10:23:19 -0500 |
| commit | ef5bd494087a46ee80d8bc17796074efdae81ff4 (patch) | |
| tree | 3104d9b8c0a07a38961aee54057125e45941db88 /experiments/confirmatory_paper_experiments.py | |
| parent | 7e01fbc0ce871857c1e1879ed0d3559e8bfae7c7 (diff) | |
Add Phase 7A: snapshot time sweep shows early snapshots have positive held-out transfer
At epoch 5 (acc=49%), Vec_M4 5-step: dL_held=-0.005 (PUR=0.70)
Oracle BP 5-step: dL_held=-0.009 (PUR=1.05)
DFA 5-step: dL_held=+0.003 (always hurts held-out)
By epoch 20, generalization window closes. Held-out failure is late-snapshot artifact.
Better credit → lower update variance (Vec=0.8 vs DFA=40), not higher.
Key implication: DFA warmup delays credit bridge past its useful window.
Credit should be used from epoch 0, not after 20% warmup.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Diffstat (limited to 'experiments/confirmatory_paper_experiments.py')
0 files changed, 0 insertions, 0 deletions