diff options
| author | YurenHao0426 <Blackhao0426@gmail.com> | 2026-03-23 19:46:08 -0500 |
|---|---|---|
| committer | YurenHao0426 <Blackhao0426@gmail.com> | 2026-03-23 19:46:08 -0500 |
| commit | 32123cb36ae9521f60c9b6f67458b931b6540ef2 (patch) | |
| tree | 4731e1dc513f5b613f80c4d20fc4114044c266d3 /models/state_bridge.py | |
| parent | bbb1a36d67f2f0c83106c1e771ea2c2fcb7fd83a (diff) | |
Add final report, plots, experiment guide, and complete NOTE.md
All experiments complete:
- Toy LQ: credit bridge matches state bridge (~0.94 costate cosine)
- CIFAR-10: credit bridge (29.6%) comparable to DFA (30.0%), both beat state bridge (18.5%)
- State bridge confirms core hypothesis: perfect state prediction != useful credit
- Terminal gradient matching is essential for credit bridge
Diffstat (limited to 'models/state_bridge.py')
0 files changed, 0 insertions, 0 deletions