78514a1673
Holistic regrades
2026-09-29 19:55:33 -04:00
ce0fae6b1c
all detectors
2026-09-29 19:34:06 -04:00
df5e8cbdd3
remove the jobs
2026-09-29 19:07:55 -04:00
cf81be9c40
stage atomic criteria
2026-09-29 18:44:17 -04:00
5f579fb5e6
fix: 2nd review changes
...
holistic, grader and atomic have been changed to be less sensitive to
crux issues and fix a few other minor things
2026-09-29 18:37:37 -04:00
757e772c94
final staging before packaging
2026-09-28 22:09:53 -04:00
6db023feef
reran all detectors
2026-09-28 18:36:24 -04:00
4d7b50e4a3
staging before detector run
2026-09-28 15:16:27 -04:00
3dffd055a4
ran ref runs
2026-09-28 15:07:38 -04:00
454c531c10
ran form, fixed issue, passed
2026-09-28 14:22:56 -04:00
a1c9c5fe95
ran coverage, fixed issue, got clear value
2026-09-28 14:13:45 -04:00
4ac67cd553
updated the 3 md files and staged them
2026-09-28 13:51:50 -04:00
8f599c6f9c
Remove git-archeology folder
2026-09-28 13:37:51 -04:00
b4f6fb977c
Lasts files
2026-09-28 13:35:56 -04:00
31e690ab35
Tried to package, had to rerun detectors
2026-09-27 07:43:59 -04:00
2b6898ab84
Tried to package, had to rerun all detectors
2026-09-27 07:28:32 -04:00
29c23a4a8f
Restore staged files step
2026-09-27 06:14:05 -04:00
a3dd68aede
ran re-grading
2026-09-27 06:12:04 -04:00
040251f69c
ran final detector before re-grading
2026-09-27 06:00:59 -04:00
f338dfe04e
ran last 2 detectors, fixed issues
2026-09-27 05:52:27 -04:00
c711a6d5b0
new holistic rubric before humanizing
2026-09-27 05:06:02 -04:00
12321a6013
det rubric coverage had problems
2026-09-27 04:54:38 -04:00
74d4534d59
re-graded 4 runs - better
2026-09-27 04:48:12 -04:00
8a63ac2209
graded 4 runs
2026-09-26 20:22:35 -04:00
5bc44d1a1e
generated atomic rubric
2026-09-26 19:39:34 -04:00
e14abf6490
copied reviewed trials
2026-09-26 19:30:23 -04:00
a15c794612
last 2 pre-trial detectors
2026-09-26 18:52:46 -04:00
e55fd1f41e
all 3 done
2026-09-26 18:50:39 -04:00
eeafd74131
2 of 3 detector issue solved
2026-09-26 18:46:28 -04:00
67217f46fc
detector again
2026-09-26 18:39:23 -04:00
620c9e2c4b
1st edits to holistic rubric and reran detectors
2026-09-26 16:37:24 -04:00
ababc68dc5
most of the pre-trial detectors - some raise issues
2026-09-26 16:30:02 -04:00
ae2dd295e2
added holistic-rubric
2026-09-26 16:09:49 -04:00
698c5bd731
authoring up, create mishandled_pro_v2
2026-09-26 15:49:09 -04:00
e55ccea018
Loaded up for the 3rd redo
...
Still on potion-voice
2026-09-26 14:57:10 -04:00
bceb52e8ee
lots of change - all to start my 3rd redo
2026-09-26 14:31:52 -04:00
7f4d388e19
removed orig worker folders
2026-09-26 13:51:06 -04:00
d54276de34
broken-dev-env detector show me git archeology never exists
...
This means I have to redo the entire project.
2026-09-26 13:47:40 -04:00
530c85f50c
after rem temp staged files step
2026-09-26 13:12:48 -04:00
6856e75265
after updates to atomic rubric, rerun detector
2026-09-26 13:11:29 -04:00
18574c0ca6
final detectors - 1 problem
2026-09-26 13:06:56 -04:00
0bd21e0d88
one regrade - no edits before hand
2026-09-25 20:22:19 -04:00
2aa9ab7ff4
fixed .gitignore, regrades
2026-09-25 20:08:25 -04:00
fe1f0de788
chmod
2026-09-25 19:25:39 -04:00
abf389d7fc
Converted mishandle_pro_v2 using the write-atomic-rubric rules:
...
Created 14 criteria in harbor-tasks/mishandle_pro_v2/tests/atomic-rubric.yaml:1.
Extracted context verbatim into harbor-tasks/mishandle_pro_v2/tests/grader-context.md:1.
Correctly used zero Crux criteria because penalties target individual dimensions.
Staging validation passed.
Form and coverage detectors both report clear with HIGH confidence.
The rubric is currently staged for grading; use --restore before packaging.
2026-09-25 16:26:24 -04:00
87b54b8a98
deleted old trails
2026-09-25 16:15:31 -04:00
e2e3f73ad5
final two detectors
2026-09-25 15:37:33 -04:00
9ec7917e12
both clear now
2026-09-25 15:33:44 -04:00
8031ff1d4c
several turns - persistenc is the issue now
2026-09-25 15:29:35 -04:00
9956c5614e
holistic-rubric fixes
...
dimmention-misapplication is fine.
rubric-clarity has an issue.
-- Verdict: material issues. The earlier ambiguities are fixed, but Ground Truth accepts “unmerged or deprecated” prototype commits while the top Thought Partnership tier requires “unmerged” commits. That difference could change the score for an accurate response.
2026-09-25 15:09:53 -04:00
