summaryrefslogtreecommitdiff
path: root/TWO_STATE_BIAS_PROGRAM.md
diff options
context:
space:
mode:
authorYurenHao0426 <Blackhao0426@gmail.com>2026-08-06 15:59:45 -0500
committerYurenHao0426 <Blackhao0426@gmail.com>2026-08-06 15:59:45 -0500
commit53f578f73929db9aaeee06ba6fe55f908e0a164f (patch)
treedfadf1606d75dc29f3c745533d04c8687c6469e6 /TWO_STATE_BIAS_PROGRAM.md
parentfc42f93cee71210532983a3962f1cd227f96e5f3 (diff)
protocol: require BP-free hardware-local debiasing
Diffstat (limited to 'TWO_STATE_BIAS_PROGRAM.md')
-rw-r--r--TWO_STATE_BIAS_PROGRAM.md6
1 files changed, 6 insertions, 0 deletions
diff --git a/TWO_STATE_BIAS_PROGRAM.md b/TWO_STATE_BIAS_PROGRAM.md
index bf5bc37..58bdf6c 100644
--- a/TWO_STATE_BIAS_PROGRAM.md
+++ b/TWO_STATE_BIAS_PROGRAM.md
@@ -52,6 +52,12 @@ backbone-specific controller and still call the result SDIL. Neutral
observations, state storage and arithmetic are charged. Baselines receive the
same observation budget when applicable.
+Every adapter is also bound by `HARDWARE_LOCALITY_CONTRACT.md`: SDIL training
+must execute with autograd disabled, and a stored local observation tuple must
+reproduce its update without the model, task loss, labels or downstream
+layers. Author code may use autograd as a native reference, but paper-facing
+SDIL runs require an audited hand-written local update.
+
## Unified bias model
For local population or edge `l`, let the clean two-state teaching estimate be