Skip to content

The full battery replicates: 53/58, twice #43

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
punnerud merged 1 commit into from
Aug 13, 2026
Merged

The full battery replicates: 53/58, twice #43

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 12 additions & 0 deletions dev/scripts/check_numbers.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,6 +48,7 @@
PHYSICS = "docs/claims/battery_physics_20260811_qwen3.json"
CALCPHYS = "docs/claims/battery_calcphys_20260811_qwen3.json"
ALL_SIX = "docs/claims/battery_all6_20260811_qwen3.json"
ALL_SIX_RUN2 = "docs/claims/battery_all6_run2_20260811_qwen3.json"

# (file, extractor, expected, tolerance, label)
#
Expand Down Expand Up @@ -525,6 +526,17 @@
0.0,
"and no run in it errored",
),
(
ALL_SIX_RUN2,
# An independent full re-run, 53/58 both times -- the replication that
# makes the pin above worth trusting. Every earlier small comparison in
# this repository swung by up to 28 points; two identical 58-question
# totals is what stability actually looks like here.
lambda d: d["correct_rate"],
0.91,
0.10,
"and a full re-run lands on the same nine in ten",
),
(
QWEN_SELECT,
# And it does the job it exists for: a value computed exactly and then not
Expand Down
Loading
Loading