qishizhe/digital-drosophila
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Audit fixed D18 credit and conditional MBON transfer (Phase 4AL)

Browse Source
This commit is contained in:
Digital Drosophila Archive
2026-09-30 13:58:39 +08:00
parent c6444255fe
commit 6d5f66bcc3
1058 changed files with 120679 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files
registry/phase4al-mb-credit-transfer/evidence/REPORT.md
+16
View File
@@ -0,0 +1,16 @@
# Phase 4AL — conditional credit/MBON transfer audit
Status: `PASS-CONDITIONAL-CREDIT-TRANSFER-AUDIT-NOT-LEARNING`.
All 120 primary trials and 24 repeats are retained. Full events/counts/final V/g/ref baseline exactness is required.
Virtual eligibility is sampled after tick 900; no DAN is injected, no delta is learned.
Uniform 25% mask depression is a temporary conditional replay; every source train, including feedback, is clamped.
All 60 odor pair-mode records and all 97 MBON targets are reported; no best pair/parameter selection.
- novel_to_aj/matched_energy: specific KC/edge sets 15/15; conditional event change 30/30; count/first-latency change 30/30.
- novel_to_aj/physiological: specific KC/edge sets 15/15; conditional event change 30/30; count/first-latency change 30/30.
- seed_validation/matched_energy: specific KC/edge sets 15/15; conditional event change 30/30; count/first-latency change 30/30.
- seed_validation/physiological: specific KC/edge sets 15/15; conditional event change 30/30; count/first-latency change 30/30.
No biological learning PASS or candidate. This does not complete the learning goal.
Read EVIDENCE.md in the model for provenance, unmatched biology and conditional/closed-loop distinctions.