anhtld commited on
Commit
c3ef28a
·
verified ·
1 Parent(s): f461a15

Auto-sync: 2026-06-29 07:26:06

Browse files
Files changed (1) hide show
  1. results/README.md +15 -3
results/README.md CHANGED
@@ -114,10 +114,12 @@ Completed deployment-clean diagnostics:
114
 
115
  Completed residual transport jobs:
116
 
117
- Last checked: `2026-06-29 10:57 UTC`. The K4 mean-by-type scale-grid sweep
118
  completed and produced a new clean best, 35.42%, while upper/wide,
119
- minimum-energy, and source-advantage follow-ups completed without improving it.
120
- The paper table/paired analysis now use that row as `best_clean_key`.
 
 
121
 
122
  - Code supports `--retrieval-residual-scale`, defaulting to `1.0` so the
123
  completed 32.12% result is unchanged.
@@ -217,6 +219,16 @@ The paper table/paired analysis now use that row as `best_clean_key`.
217
  no-op 0.03 reaches 34.43%, and K2 safe residual retrieval reaches 34.26%;
218
  both are below the raw-metric rows at 35.25% and 35.01%. Summary jobs
219
  `14893788`/`14893790` and rebuild job `14893791` completed.
 
 
 
 
 
 
 
 
 
 
220
  - Job `14894281` completed the unit smoke for the source-progress viability gate,
221
  including the variable residual-count padding check.
222
  - Job `14894282` completed the ManiSkill CPU rollout smoke for K4 mean-by-type
 
114
 
115
  Completed residual transport jobs:
116
 
117
+ Last checked: `2026-06-29 11:25 UTC`. The K4 mean-by-type scale-grid sweep
118
  completed and produced a new clean best, 35.42%, while upper/wide,
119
+ minimum-energy, source-advantage, and train-family success-prior follow-ups
120
+ completed without improving it. Consensus-confidence jobs are running/pending
121
+ as the next geometric reliability test. The paper table/paired analysis now use
122
+ the scale-grid no-op row as `best_clean_key` unless those jobs improve it.
123
 
124
  - Code supports `--retrieval-residual-scale`, defaulting to `1.0` so the
125
  completed 32.12% result is unchanged.
 
219
  no-op 0.03 reaches 34.43%, and K2 safe residual retrieval reaches 34.26%;
220
  both are below the raw-metric rows at 35.25% and 35.01%. Summary jobs
221
  `14893788`/`14893790` and rebuild job `14893791` completed.
222
+ - Jobs `14903128`/`14903130`/`14903132`/`14903134` completed the continuous
223
+ train-family success-prior sweep. Family-success bonuses `0.02`, `0.03`, and
224
+ `0.05` reach 35.25%; adding family-success `0.02` to no-op `0.03` ties the
225
+ 35.42% scale-grid best without improving it. Summary jobs `14903129`/
226
+ `14903131`/`14903133`/`14903135` and rebuild job `14903136` completed.
227
+ - Job `14903296` completed the CPU smoke for train-neighbor
228
+ consensus-confidence penalties. Jobs `14903384`/`14903386`/`14903388`/
229
+ `14903390` are the submitted GPU arrays for consensus-only `0.05` and no-op
230
+ `0.03` plus consensus penalties `0.02`, `0.05`, and `0.10`; summaries are
231
+ `14903385`/`14903387`/`14903389`/`14903391`, with rebuild `14903392`.
232
  - Job `14894281` completed the unit smoke for the source-progress viability gate,
233
  including the variable residual-count padding check.
234
  - Job `14894282` completed the ManiSkill CPU rollout smoke for K4 mean-by-type