Rework the scoring presentation: per-outcome counts, colors, collapsible details - #172
Merged
Conversation
The checker table is sorted by missed rejections first and missed acceptances second, so show the ✋ column before the 👍 column to make that sort order obvious. Reorder the stat cards on the checker pages and the explanatory text below the table to match. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0128eW9GdRX7bhEnjfMhMkbW
Instead of 23/42-style ratios, group the columns by expected outcome (✋ reject, 👍 accept, 🤷 either) under a two-row header, and count in the second row what the checker did: correct, wrong (not applicable for 🤷) and declined or crashed. Zero counts are left blank, and a non-zero wrong count puts a red background on just that cell. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0128eW9GdRX7bhEnjfMhMkbW
The table is sorted by wrongly accepted invalid proofs, wrongly rejected valid proofs, mathlib checking time (checkers that do not manage mathlib below those that do), and the number of declined tests, in that order; the details section previously only mentioned the first two criteria. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0128eW9GdRX7bhEnjfMhMkbW
Use ✅/❌ in the second header row, which counts correct and wrong answers, instead of repurposing the accept/reject emojis. Keep header rows out of chota's zebra striping, which greyed the second header row, and shrink the header padding so the two rows don't tower over the table. Mark non-zero wrong cells with the light bg-danger red rather than the saturated bg-error. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0128eW9GdRX7bhEnjfMhMkbW
The round indicator line under the title was misplaced; the index page title now reads "Lean Kernel Arena – Round <name>", or "– Round in progress", directly in the header. The link to the other rounds is dropped from the header for now; the Rounds prose still links there. The Details section moves before the checker table, becomes collapsible without JavaScript via a <details> element, collapsed by default, and is broken into subsections (Scoring, Performance, Downloads, Contributing, Rounds), with the former separate Rounds section becoming the last subsection. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0128eW9GdRX7bhEnjfMhMkbW
Wrongly accepting an invalid proof (unsound) is now marked bg-error, and wrongly rejecting a valid proof (incomplete) bg-warning — in the checker summary, the tests matrix, the group summary rows and cells, and the checker pages. Declined or crashed tests are no longer colored; they are excluded from the scoring anyway. Group summaries track the two wrong directions separately, also in their tooltips. The light palette in the per-page styles never actually applied to .bg-error and .bg-success: chota's saturated definitions carry !important and won. The palette now lives in style.css with matching !important, so every page gets the intended light colors. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0128eW9GdRX7bhEnjfMhMkbW
Instead of distinguishing the direction of a wrong answer by color, mark every correct answer green, every wrong one red, and every declined or crashed test yellow — in the checker summary, the tests matrix, the group summaries, and the Result column on the per-checker pages. Group cells color by the worst result present: red before yellow before green. On perf-compared cells the performance highlight still takes precedence over the plain green. The per-direction counts remain in the group tooltips. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0128eW9GdRX7bhEnjfMhMkbW
The soundness and completeness cards now show the same three numbers as the main index, with the same icons and tooltips: correct ✅, wrong ❌ and declined 🚫, instead of a 64/65-style ratio that left the declines implicit. The separate declined-tests card is replaced by a card for the corner case tests, which have no prescribed outcome and only distinguish processed from declined. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0128eW9GdRX7bhEnjfMhMkbW
In the tests table green is the performance signal (fastest checker, faster than official), so a merely correct result stays uncolored there, in the cells and in the group summary rows. The checker summary table and the checker and test pages keep the green for correct. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0128eW9GdRX7bhEnjfMhMkbW
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Reworks how results are presented, starting from "put the soundness column
first" and iterating from there:
groups the columns by expected outcome (✋ reject, 👍 accept, 🤷 either) and
counts what the checker did (✅ correct, ❌ wrong, 🚫 declined/crashed).
Zero counts render as blank cells. Soundness comes before completeness,
matching the sort order, which the Details section now states in full
(wrongly accepted, wrongly rejected, mathlib time, declines).
the summary table, group summaries, and the checker/test pages — except in
the tests table, where green stays reserved for the perf signal. The light
palette now actually applies: the intended light colors were being
overridden by chota's
!importantrules.the index (
64 ✅ 0 ❌ 1 🚫); the declined card is replaced by a cornercases card. Group tooltips distinguish wrongly accepted from wrongly
rejected.
Round in progress"), replacing the round line under it. The Details section
moved above the tables, is collapsible without JavaScript (
<details>),collapsed by default, and split into subsections, with Rounds as one of
them.
🤖 Generated with Claude Code
https://claude.ai/code/session_0128eW9GdRX7bhEnjfMhMkbW