The swe-bench leaderboard can no longer order its top entries. a september 15 audit shows the gaps…
the swe-bench leaderboard can no longer order its top entries. a september 15 audit shows the gaps at the top are smaller than the noise between runs, so the ranking reads as confidence where there is none.
benchmarks stopped separating agents.
Context
The arXiv paper 2609.17394, dated 15 September 2026 and accepted at ADMA 2026 per its comments field, audits 254 SWE-bench submissions across four splits without running models. On Verified the leading two entries each resolve 396 of 500 instances, the top ten share 285 successes and 51 failures leaving 164 instances that distinguish them, and exact paired McNemar tests separate none of the 29 adjacent top-thirty pairs at alpha 0.05, while the larger Test split separates 14 of 23. Within-model scaffold ranges reach 29.8 percentage points against an 8.8-point spread of the top thirty. The paper says non-rejection does not establish equivalence and that the observational design does not identify causal scaffold effects.
The audit works on published per-instance outcomes and ran no models, so it did not measure noise between runs; smaller than the noise between runs is not what it measured, and it tests paired instance-level differences. The result is specific to the Verified top thirty, and on the larger Test split 14 of 23 pairs separate. Not separating adjacent pairs is not equivalence. Benchmarks stopped separating agents generalises beyond the paper and is the author's thesis.
Watch next
- Whether SWE-bench adds confidence intervals and a peer-reviewed version.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 20 September 2026 at 11:46 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →