Auto-sync: 2026-06-28 17:43:12

results/README.md

@@ -91,8 +91,9 @@ Completed deployment-clean diagnostics:
 
 Completed residual transport jobs:
 
-Last checked: `2026-06-28 17:54 UTC`. The counterfactual tangent ray-search batch
-completed and table/analysis outputs were rebuilt.
+Last checked: `2026-06-28 21:52 UTC`. The counterfactual tangent ray-search batch
+completed; the typed no-op residual-prior clean sweep also completed after a
+passing CPU smoke.
 
 - Code supports `--retrieval-residual-scale`, defaulting to `1.0` so the
   completed 32.12% result is unchanged.
@@ -118,8 +119,8 @@ completed and table/analysis outputs were rebuilt.
   diagnostics. Thresholds through `0.75` do not filter enough; scale `0.50`
   stays at 33.33%, and scale `0.25`/threshold `0.25` is 32.93%.
 - Jobs `14859503`-`14859597` completed typed-safe residual scale fine/zoom
-  sweeps. Scales `0.325`, `0.35`, and `0.40` tie at the current clean best,
-  33.74%; scale `0.35` has the lowest seed std among the tied rows.
+  sweeps. Scales `0.325`, `0.35`, and `0.40` tie at the pre-abstention clean
+  best, 33.74%; scale `0.35` has the lowest seed std among the tied rows.
 - Jobs `14862455`-`14862460` completed residual-tangent target export,
   distillation, and direct/best-rank rollouts. The tangent student is a negative
   diagnostic: 28.87% for best-policy rollout.
@@ -129,7 +130,7 @@ completed and table/analysis outputs were rebuilt.
   33.33%/33.28%.
 - Jobs `14862635`-`14862828` completed counterfactual-advantage margin sweeps.
   Scale `0.35`, safe residual families, margin `0.20` or `0.22` reaches 34.84%.
-- Jobs `14862857`-`14862939` completed KNN-with-abstention sweeps. The current
+- Jobs `14862857`-`14862939` completed KNN-with-abstention sweeps. The previous
   clean best is K2 retrieval, scale `0.40`, safe residual families, margin
   `0.20`: 35.01%.
 - Jobs `14868661`-`14868668` completed same-state no-expert lattice with a
@@ -137,15 +138,22 @@ completed and table/analysis outputs were rebuilt.
   56.99% no-expert lattice, so policy fallback is a negative mechanism diagnostic.
 - Jobs `14868693`-`14868700` completed the clean mean-by-type tangent consensus
   sweep. K4, scale `0.40`, margin `0.20` reaches 34.96%, close to but below the
-  35.01% K2 raw residual best.
+  35.01% K2 raw residual best before typed no-op priors.
 - Jobs `14868798`-`14868805` completed the consensus follow-up. K4 mean scales
   `0.425` and `0.45` reach 34.72% and 34.84%; K4 median and K8 mean at scale
-  `0.40` both reach 34.67%. K2 raw residual transport remains the best clean row.
+  `0.40` both reach 34.67%.
 - Jobs `14868993`/`14868995`/`14868997`/`14868999` completed the counterfactual
   tangent ray-search batch. Results are 34.84% for K1 tight, 34.84% for K2
-  tight, 34.96% for K2 broad, and 34.55% for K4 tight, so fixed K2 residual
-  transport at 35.01% remains the best clean row. Summary jobs `14868994`/
-  `14868996`/`14868998`/`14869000` and rebuild job `14869860` completed.
+  tight, 34.96% for K2 broad, and 34.55% for K4 tight. Summary jobs
+  `14868994`/`14868996`/`14868998`/`14869000` and rebuild job `14869860`
+  completed.
+- Job `14883591` completed the CPU smoke for candidate-type potential bonuses
+  with `residual_no_op=0.05`, confirming the new CLI/Slurm path.
+- Jobs `14883919`/`14883921`/`14883923` completed GPU clean sweeps for K4
+  mean-by-type residual retrieval with no-op residual bonuses `0.03`, `0.05`,
+  and `0.08`. Results are 35.25%, 35.19%, and 35.13%; bonus `0.03` is the
+  current best clean deployment diagnostic. Summary jobs `14883920`/`14883922`/
+  `14883924` and rebuild job `14883926` completed.
 - Job `14869627` completed the CPU Apptainer smoke for the residual scale-grid
   selector: selected index `3` and returned action `0.20` on the toy ray-search
   case, validating candidate expansion before rollout.