diff options
| author | YurenHao0426 <Blackhao0426@gmail.com> | 2026-07-22 13:22:24 -0500 |
|---|---|---|
| committer | YurenHao0426 <Blackhao0426@gmail.com> | 2026-07-22 13:22:24 -0500 |
| commit | c24a6a575bce15e7d7da541ecb752c65b5e68e47 (patch) | |
| tree | 8166e8c7eebf2785cd811e4fbbc1578b185262bc /BASELINES.md | |
| parent | 225459f7ddfcd3c8d89f0515a0d766bf2a15c848 (diff) | |
results: close fixed hierarchical FA short gate
Diffstat (limited to 'BASELINES.md')
| -rw-r--r-- | BASELINES.md | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/BASELINES.md b/BASELINES.md index 9800d16..6b76d09 100644 --- a/BASELINES.md +++ b/BASELINES.md @@ -20,7 +20,8 @@ The audited implementations and completed results are in `RESULTS.md`: adjoints, and local ReLU/BatchNorm Jacobians; it never copies or reads a downstream forward convolution and is the mandatory baseline for any learned hierarchical extension. Its frozen validation-only screen is specified in - `HFA_BASELINE.md`; + `HFA_BASELINE.md`; the selected short endpoint is 43.52% versus matched DFA's + 37.16%, but its 50% full-run gate failed; - direct node perturbation uses a causal target on every hidden update and no learned vectorizer; - the no-traffic, `P=0` SDIL backbone is learned direct feedback by node |
