Bugfix · Issue Resolution (SWE-bench-style)

SWE-bench-Pro-style graded scoring: each arm fixes a buggy repo from the issue. Resolved = the shown hidden test passes (FAIL_TO_PASS). Strict also requires held-out anti-overfit tests + PASS_TO_PASS regression tests (the arm never sees them) with no test-file edits. The composite [0–100] grades robustness, regression, patch minimality/locality and code health — so two arms that both "resolve" still separate.

Issue resolution, held-out checks, and patch inspection.

The broad historical reference runs, not the later one-case diagnostics. These use modeled/mock generation and are not empirical evidence of current agent performance. Each track retains its own date, model labels and scoring history.

The retained full modeled matrix covers six issues. The captured collection separately covers eleven issues.

Models & settings — click to expand

Patch-quality composite by harness

Graded quality-of-fix [0–100]: robustness (held-out) 40% · regression 25% · minimality 15% · locality 10% · code health 10%, gated on resolution (blended 80/20 with the semantic patch judge when one is configured — the live default)

Resolution: shown vs strict (held-out + regression)

Lenient = shown hidden test passes. Strict = held-out anti-overfit + PASS_TO_PASS regression also pass (no test edits). A gap = overfit/regressing patches.

Resolution explorer

Per issue × harness: composite, strict resolution, held-out (anti-overfit) + regression pass counts, patch minimality, and the patched files.
IssueHarnessCompositeStrictHiddenHeld-outRegressionMinimalityBackend

Inspect the evidence

Inside the dataset

6 items
sum-evenssum_evens returns the sum of odd numbers
mathlib.calc.sum_evens(nums) should return the sum of the EVEN numbers in nums, but it currently returns the sum of the odd numbers. Fix it so even numbers are summed (and an empty list sums to 0).
Open scored results ↓
stack-lifoStack.pop removes the oldest item (FIFO) instead of the newest (LIFO)
ds.stack.Stack is meant to be a last-in-first-out stack, but Stack.pop() removes the OLDEST item instead of the most recently pushed one. Fix pop() so the stack is LIFO; popping an empty stack should raise IndexError.
Open scored results ↓
slugify-normalizeslugify does not lowercase or strip punctuation
textutil.slug.slugify(s) should produce a URL slug: lowercase, words separated by single hyphens, with all non-alphanumeric characters removed and no leading/trailing hyphens. It currently only replaces spaces with hyphens. Fix it to fully normalize the string.
Open scored results ↓
paginate-ceilpage_count truncates the final partial page
pagelib.page.page_count(total, per_page) should return the number of pages needed to show `total` items at `per_page` per page, rounding up for a partial last page. It currently uses integer division, dropping the final partial page. Fix it (0 items is 0 pages).
Open scored results ↓
wordcount-casefoldword_count is case-sensitive
countlib.words.word_count(text) should count word frequencies case-insensitively, returning a dict of lowercase word -> count. It currently counts case-sensitively, so 'The' and 'the' are separate keys. Fix it to fold case.
Open scored results ↓
clamp-boundsclamp ignores the lower bound
numlib.clamp.clamp(x, lo, hi) should constrain x to the inclusive range [lo, hi]. It currently applies only the upper bound, so values below lo pass through unchanged. Fix it to apply both.
Open scored results ↓

Inspect the evidence

Explore experiments

21 items
2026-06-22 · Historical run11 cases × 5 harnessestui-repo-20260622-184430-368Plots, evidence & explorer →2026-06-22 · Historical run6 cases × 5 harnessestui-repo-20260620-183114-752-rescoredPlots, evidence & explorer →2026-06-21 · Historical run6 cases × 5 harnessestui-repo-20260620-183114-752Plots, evidence & explorer →2026-06-19 · Historical run6 cases × 5 harnessesrepo-20260619-085409-218Plots, evidence & explorer →2026-06-18 · Historical run6 cases × 5 harnessestui-repo-20260618-161843-254Plots, evidence & explorer →2026-06-18 · Historical run6 cases × 5 harnessestui-repo-20260618-143819-265Plots, evidence & explorer →2026-06-18 · Historical run6 cases × 5 harnessestui-repo-20260618-143655-650Plots, evidence & explorer →2026-06-18 · Historical run6 cases × 5 harnessesrepo-20260618-124503-769Plots, evidence & explorer →2026-06-18 · Historical run6 cases × 5 harnessesrepo-20260618-124426-652Plots, evidence & explorer →2026-06-18 · Historical run6 cases × 5 harnessesrepo-20260618-124352-258Plots, evidence & explorer →2026-06-18 · Historical run6 cases × 5 harnessesrepo-20260618-114225-007Plots, evidence & explorer →2026-06-18 · Historical run6 cases × 5 harnessesrepo-20260618-113532-802Plots, evidence & explorer →2026-06-18 · Historical run6 cases × 5 harnessesrepo-20260618-112706-494Plots, evidence & explorer →2026-06-18 · Historical run6 cases × 5 harnessesrepo-20260618-112213-706Plots, evidence & explorer →2026-06-18 · Historical run6 cases × 1 harnessesrepo-20260618-110728-006Plots, evidence & explorer →2026-06-18 · Historical run6 cases × 5 harnessesrepo-20260618-004613-261Plots, evidence & explorer →2026-06-23 · Historical run1 cases × 1 harnessesrepo-20260623-132643-640Plots, evidence & explorer →2026-06-18 · Historical run1 cases × 3 harnessesrepo-20260618-114816-279Plots, evidence & explorer →2026-06-18 · Historical run1 cases × 1 harnessesrepo-20260618-110430-432Plots, evidence & explorer →2026-06-18 · Historical run1 cases × 5 harnessesrepo-20260618-014254-590Plots, evidence & explorer →2026-06-22 · Historical run0 cases × 0 harnessestui-repo-20260622-184430-368-rescoredPlots, evidence & explorer →