Update experiment measurements and status
Browse files- reports/eval/control-2027-1000.json +0 -0
- reports/selection.json +116 -0
- reports/status.json +4 -4
reports/eval/control-2027-1000.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
reports/selection.json
ADDED
|
@@ -0,0 +1,116 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"objective": "rank_1_efficiency_after_full_confirmation",
|
| 3 |
+
"selection_version": "combined-development-v2-before-pilots",
|
| 4 |
+
"results": [
|
| 5 |
+
{
|
| 6 |
+
"source": "fineweb_edu",
|
| 7 |
+
"effects": [
|
| 8 |
+
{
|
| 9 |
+
"seed": 1337,
|
| 10 |
+
"arc_delta_pp": 0.17543859649122806,
|
| 11 |
+
"grammar_delta_pp": 2.430555555555556,
|
| 12 |
+
"wiki_ppl_ratio": 0.9998499487147989,
|
| 13 |
+
"wiki_normalized_delta": 0.0026825473015649644,
|
| 14 |
+
"combined_dev_delta": 0.869558899782783
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"seed": 2027,
|
| 18 |
+
"arc_delta_pp": 0.3508771929824561,
|
| 19 |
+
"grammar_delta_pp": 0.6944444444444444,
|
| 20 |
+
"wiki_ppl_ratio": 0.9999204588938312,
|
| 21 |
+
"wiki_normalized_delta": 0.001421948878126593,
|
| 22 |
+
"combined_dev_delta": 0.3489145287683424
|
| 23 |
+
}
|
| 24 |
+
],
|
| 25 |
+
"mean_combined_dev_delta": 0.6092367142755627,
|
| 26 |
+
"paired_dev_99pct_interval": [
|
| 27 |
+
-0.2904952545353535,
|
| 28 |
+
1.7977113731449583
|
| 29 |
+
],
|
| 30 |
+
"eligible_for_extension": false
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"source": "dclm_edu",
|
| 34 |
+
"effects": [
|
| 35 |
+
{
|
| 36 |
+
"seed": 1337,
|
| 37 |
+
"arc_delta_pp": 0.17543859649122806,
|
| 38 |
+
"grammar_delta_pp": 1.3888888888888888,
|
| 39 |
+
"wiki_ppl_ratio": 0.9999958367943211,
|
| 40 |
+
"wiki_normalized_delta": 7.442243152627325e-05,
|
| 41 |
+
"combined_dev_delta": 0.5214673026038811
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"seed": 2027,
|
| 45 |
+
"arc_delta_pp": 0.3508771929824561,
|
| 46 |
+
"grammar_delta_pp": 0.0,
|
| 47 |
+
"wiki_ppl_ratio": 0.9998027158537773,
|
| 48 |
+
"wiki_normalized_delta": 0.003527037797255739,
|
| 49 |
+
"combined_dev_delta": 0.11813474359323728
|
| 50 |
+
}
|
| 51 |
+
],
|
| 52 |
+
"mean_combined_dev_delta": 0.3198010230985592,
|
| 53 |
+
"paired_dev_99pct_interval": [
|
| 54 |
+
-0.40827828976693786,
|
| 55 |
+
1.1902993174455379
|
| 56 |
+
],
|
| 57 |
+
"eligible_for_extension": false
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"source": "cosmopedia",
|
| 61 |
+
"effects": [
|
| 62 |
+
{
|
| 63 |
+
"seed": 1337,
|
| 64 |
+
"arc_delta_pp": 0.3508771929824561,
|
| 65 |
+
"grammar_delta_pp": 1.7361111111111112,
|
| 66 |
+
"wiki_ppl_ratio": 1.0022786467949207,
|
| 67 |
+
"wiki_normalized_delta": -0.04068719618935495,
|
| 68 |
+
"combined_dev_delta": 0.6821003693014039
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"seed": 2027,
|
| 72 |
+
"arc_delta_pp": -0.8771929824561403,
|
| 73 |
+
"grammar_delta_pp": 0.3472222222222222,
|
| 74 |
+
"wiki_ppl_ratio": 1.0025389677110237,
|
| 75 |
+
"wiki_normalized_delta": -0.04532956111627206,
|
| 76 |
+
"combined_dev_delta": -0.1917667737833967
|
| 77 |
+
}
|
| 78 |
+
],
|
| 79 |
+
"mean_combined_dev_delta": 0.24516679775900363,
|
| 80 |
+
"paired_dev_99pct_interval": [
|
| 81 |
+
-0.7168184605255966,
|
| 82 |
+
1.2868486934705048
|
| 83 |
+
],
|
| 84 |
+
"eligible_for_extension": false
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"source": "finemath",
|
| 88 |
+
"effects": [
|
| 89 |
+
{
|
| 90 |
+
"seed": 1337,
|
| 91 |
+
"arc_delta_pp": -0.17543859649122806,
|
| 92 |
+
"grammar_delta_pp": 2.083333333333333,
|
| 93 |
+
"wiki_ppl_ratio": 0.9999358419833166,
|
| 94 |
+
"wiki_normalized_delta": 0.0011469380111162695,
|
| 95 |
+
"combined_dev_delta": 0.6363472249510737
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"seed": 2027,
|
| 99 |
+
"arc_delta_pp": 0.0,
|
| 100 |
+
"grammar_delta_pp": 0.6944444444444444,
|
| 101 |
+
"wiki_ppl_ratio": 0.9999406018055441,
|
| 102 |
+
"wiki_normalized_delta": 0.0010618452438650554,
|
| 103 |
+
"combined_dev_delta": 0.23183542989610317
|
| 104 |
+
}
|
| 105 |
+
],
|
| 106 |
+
"mean_combined_dev_delta": 0.4340913274235884,
|
| 107 |
+
"paired_dev_99pct_interval": [
|
| 108 |
+
-0.26400808778108986,
|
| 109 |
+
1.3345603820044878
|
| 110 |
+
],
|
| 111 |
+
"eligible_for_extension": false
|
| 112 |
+
}
|
| 113 |
+
],
|
| 114 |
+
"selected_source": null,
|
| 115 |
+
"note": "Development-only surrogate; narrow grammar proxy; interval conditional on fixed WikiText validation corpus. Two seeds do not establish population variance. No official GLINT scores."
|
| 116 |
+
}
|
reports/status.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
| 1 |
{
|
| 2 |
-
"stage": "
|
| 3 |
-
"utc": "2026-09-30T23:
|
| 4 |
-
"remaining_seconds":
|
| 5 |
-
"
|
| 6 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"stage": "complete_no_reliable_winner",
|
| 3 |
+
"utc": "2026-09-30T23:41:21.495137+00:00",
|
| 4 |
+
"remaining_seconds": 30452.50485,
|
| 5 |
+
"message": "No source passed the paired development gate; no 5B run launched."
|
| 6 |
}
|