anhtld commited on
Commit
62a37d8
·
verified ·
1 Parent(s): 81fee44

Auto-sync: 2026-06-29 01:53:54

Browse files
Files changed (1) hide show
  1. results/README.md +11 -2
results/README.md CHANGED
@@ -46,6 +46,9 @@ Use these summaries for the current paper tables:
46
  `h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_noopbonus0p03_srcprog0p75_summary.md`,
47
  `h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_srcprogbonus0p03_summary.md`,
48
  `h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_srcprogbonus0p05_summary.md`,
 
 
 
49
  `h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_fieldsoftmax_grid_safe_margin0p20_summary.md`,
50
  `h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_fieldsoftmax_grid_safe_margin0p10_noopbonus0p03_summary.md`,
51
  `h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_policy_anchor_scale0p35_safe_noexpert_summary.md`,
@@ -98,9 +101,9 @@ Completed deployment-clean diagnostics:
98
 
99
  Completed residual transport jobs:
100
 
101
- Last checked: `2026-06-29 03:58 UTC`. The source-progress bonus-prior batch
102
  completed after passing CPU/unit smokes, and the paper table/paired analysis now
103
- include both threshold gates and continuous train-source progress priors.
104
 
105
  - Code supports `--retrieval-residual-scale`, defaulting to `1.0` so the
106
  completed 32.12% result is unchanged.
@@ -217,6 +220,12 @@ include both threshold gates and continuous train-source progress priors.
217
  no-op prior. Bonus `0.03` ties the current best at 35.25%, while bonus `0.05`
218
  reaches 35.13%. Summary jobs `14894676`/`14894677` completed; rebuild job
219
  `14894678` was queued after them.
 
 
 
 
 
 
220
  - Job `14869627` completed the CPU Apptainer smoke for the residual scale-grid
221
  selector: selected index `3` and returned action `0.20` on the toy ray-search
222
  case, validating candidate expansion before rollout.
 
46
  `h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_noopbonus0p03_srcprog0p75_summary.md`,
47
  `h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_srcprogbonus0p03_summary.md`,
48
  `h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_srcprogbonus0p05_summary.md`,
49
+ `h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_srcscorebonus0p015_summary.md`,
50
+ `h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_srcscorebonus0p02_summary.md`,
51
+ `h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4s040_safe_margin0p20_mean_by_type_srcscorebonus0p025_summary.md`,
52
  `h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_fieldsoftmax_grid_safe_margin0p20_summary.md`,
53
  `h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_k4_fieldsoftmax_grid_safe_margin0p10_noopbonus0p03_summary.md`,
54
  `h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_policy_anchor_scale0p35_safe_noexpert_summary.md`,
 
101
 
102
  Completed residual transport jobs:
103
 
104
+ Last checked: `2026-06-29 05:54 UTC`. The source-score bonus-prior batch
105
  completed after passing CPU/unit smokes, and the paper table/paired analysis now
106
+ include progress and reward-score train-outcome priors.
107
 
108
  - Code supports `--retrieval-residual-scale`, defaulting to `1.0` so the
109
  completed 32.12% result is unchanged.
 
220
  no-op prior. Bonus `0.03` ties the current best at 35.25%, while bonus `0.05`
221
  reaches 35.13%. Summary jobs `14894676`/`14894677` completed; rebuild job
222
  `14894678` was queued after them.
223
+ - Jobs `14897121`/`14897122` completed unit and CPU rollout smokes for the
224
+ train-source reward-score bonus path.
225
+ - Jobs `14897123`/`14897124`/`14897125` completed source-score bonus arrays.
226
+ Bonuses `0.015` and `0.020` tie the current best at 35.25%, while bonus
227
+ `0.025` reaches 35.19%. Summary jobs `14897126`/`14897127`/`14897128` and
228
+ rebuild job `14897129` completed.
229
  - Job `14869627` completed the CPU Apptainer smoke for the residual scale-grid
230
  selector: selected index `3` and returned action `0.20` on the toy ray-search
231
  case, validating candidate expansion before rollout.