Kusshal commited on
Commit
6ac4226
·
verified ·
1 Parent(s): db5c75b

sync EmbedEd artifacts (phase01-v8)

Browse files
artifacts/phase4_decisions.json CHANGED
@@ -1,5 +1,5 @@
1
  {
2
- "recorded_utc": "2026-10-21 04:44 UTC",
3
  "version": "phase01-v8",
4
  "decisions": {
5
  "D4-1": {
 
1
  {
2
+ "recorded_utc": "2026-10-21 05:11 UTC",
3
  "version": "phase01-v8",
4
  "decisions": {
5
  "D4-1": {
report/measurements.md CHANGED
@@ -777,7 +777,7 @@ Phase 2 checkpoints (C0 = base model) scored on the same s′ partition without
777
 
778
  ## Phase 4 — the one table
779
 
780
- Assembled from the run dirs by the Phase-4 notebook — version `phase01-v8`, recorded 2026-10-21 04:54 UTC. Sources: `runs/<run>/eval_metrics.json`, the audit sections below, and the s′ table. **The two panels are different rulers** (benchmark: natural 13.7% prevalence, all-negative ceiling 0.2406; s′: balanced slices) — absolute F1 is NOT comparable across panels; within a panel, orderings are the finding.
781
 
782
  ### Panel A — benchmark (F1 mean ± sd over seeds 13-15; pinned = validation thresholds saturated at 1.0)
783
 
@@ -810,3 +810,20 @@ Assembled from the run dirs by the Phase-4 notebook — version `phase01-v8`, re
810
  ### Panel C — mechanism bookkeeping
811
 
812
  Trigger = an FN that is simultaneously a functional clone and a J ≥ 0.40 near-duplicate; rate = dose; margin 0.10 = amplifier (§9.4 FINAL). The 40% cut removes ~40% of C2 contamination and 0 audited clean pairs. audit_c4 PASSED (30% ≤ 42%); audit_c5/c6 outcomes: see section 3 — no C5/C6 headline before its branch is recorded.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
777
 
778
  ## Phase 4 — the one table
779
 
780
+ Assembled from the run dirs by the Phase-4 notebook — version `phase01-v8`, recorded 2026-10-21 05:11 UTC. Sources: `runs/<run>/eval_metrics.json`, the audit sections below, and the s′ table. **The two panels are different rulers** (benchmark: natural 13.7% prevalence, all-negative ceiling 0.2406; s′: balanced slices) — absolute F1 is NOT comparable across panels; within a panel, orderings are the finding.
781
 
782
  ### Panel A — benchmark (F1 mean ± sd over seeds 13-15; pinned = validation thresholds saturated at 1.0)
783
 
 
810
  ### Panel C — mechanism bookkeeping
811
 
812
  Trigger = an FN that is simultaneously a functional clone and a J ≥ 0.40 near-duplicate; rate = dose; margin 0.10 = amplifier (§9.4 FINAL). The 40% cut removes ~40% of C2 contamination and 0 audited clean pairs. audit_c4 PASSED (30% ≤ 42%); audit_c5/c6 outcomes: see section 3 — no C5/C6 headline before its branch is recorded.
813
+
814
+ ## False-negative audit — C5 (blind; SCOPE P2-9)
815
+
816
+ `python -m scripts.audit_sample score` — version `phase01-v8`, conditions C5. Share of mined negatives judged to be clones (functional equivalence). Headline counts `unsure` as not_clone; the upper bound counts it as clone.
817
+
818
+ | condition | n | clone | not_clone | unsure | FN rate [95% CI] | upper bound incl. unsure |
819
+ |---|---|---|---|---|---|---|
820
+ | C5 | 50 | 15 | 34 | 1 | **30%** [19%, 44%] | 32% [21%, 46%] |
821
+
822
+ ## False-negative audit — C6 (blind; SCOPE P2-9)
823
+
824
+ `python -m scripts.audit_sample score` — version `phase01-v8`, conditions C6. Share of mined negatives judged to be clones (functional equivalence). Headline counts `unsure` as not_clone; the upper bound counts it as clone.
825
+
826
+ | condition | n | clone | not_clone | unsure | FN rate [95% CI] | upper bound incl. unsure |
827
+ |---|---|---|---|---|---|---|
828
+ | C6 | 50 | 15 | 34 | 1 | **30%** [19%, 44%] | 32% [21%, 46%] |
829
+
sync_manifest.json CHANGED
@@ -1,6 +1,6 @@
1
  {
2
  "version": "phase01-v8",
3
- "synced_at_utc": "2026-10-21T05:10:43Z",
4
  "staged_files": [
5
  "artifacts/audit/audit_key.csv",
6
  "artifacts/audit/audit_labels.csv",
@@ -27,6 +27,7 @@
27
  "artifacts/pairs_train.tsv",
28
  "artifacts/pairs_valid.tsv",
29
  "artifacts/phase3_decisions.json",
 
30
  "artifacts/runs/C0_13/eval_metrics.json",
31
  "artifacts/runs/C1_13/eval_metrics.json",
32
  "artifacts/runs/C1_13/metrics.json",
@@ -298,7 +299,8 @@
298
  "artifacts/triples_C3.jsonl",
299
  "artifacts/triples_C4.jsonl",
300
  "artifacts/triples_C5.jsonl",
301
- "artifacts/triples_C6.jsonl"
 
302
  ],
303
  "excluded_patterns": [
304
  "sprime/*",
 
1
  {
2
  "version": "phase01-v8",
3
+ "synced_at_utc": "2026-10-21T05:11:16Z",
4
  "staged_files": [
5
  "artifacts/audit/audit_key.csv",
6
  "artifacts/audit/audit_labels.csv",
 
27
  "artifacts/pairs_train.tsv",
28
  "artifacts/pairs_valid.tsv",
29
  "artifacts/phase3_decisions.json",
30
+ "artifacts/phase4_decisions.json",
31
  "artifacts/runs/C0_13/eval_metrics.json",
32
  "artifacts/runs/C1_13/eval_metrics.json",
33
  "artifacts/runs/C1_13/metrics.json",
 
299
  "artifacts/triples_C3.jsonl",
300
  "artifacts/triples_C4.jsonl",
301
  "artifacts/triples_C5.jsonl",
302
+ "artifacts/triples_C6.jsonl",
303
+ "report/measurements.md"
304
  ],
305
  "excluded_patterns": [
306
  "sprime/*",