The benchmark just started auditing itself. testjack, out oct 7, regenerates tests from prompt…
the benchmark just started auditing itself. testjack, out oct 7, regenerates tests from prompt requirements to catch patches that pass fixed tests but break the ask.
the leaderboard score now has a witness.
Context
Verified source: the arXiv paper 'TestJack: Should you trust the results in coding benchmarks?' (arXiv 2610.10619) audits coding benchmarks by generating dynamic tests to catch reward hacking, arguing that passing a fixed set of unit tests only checks part of what a task asked for.
canberk.me (Oct 9, 2026) summarizes a preprint from Columbia, UC Berkeley and George Mason, dated arXiv Oct 7: of 4,487 passing trials across five benchmarks, TestJack refuted 1,545 (34.4 percent) and lowered the overall resolution rate from 50.6 to 33.2 percent.
The Oct 7 date and the idea of regenerating tests to catch patches that pass fixed tests match the paper and canberk.me's summary. The exact method 'regenerates tests from prompt requirements' was described as generating dynamic tests, so the wording is close but not the paper's. The 34.4 percent refutation figure is canberk.me's summary, which also lists caveats. 'The leaderboard score now has a witness' is the author's.
'the benchmark just started auditing itself' is the author's opinion.
Related work
Watch next
- Read the paper's section on which benchmarks had the highest refutation rate.
Sources
- arXiv, TestJack (2610.10619)arxiv.org
- arXiv Signals, TestJack, Oct 9, 2026arxivsignals.io
- canberk.me, Oct 9, 2026canberk.me
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 11 October 2026 at 10:49 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 →