Unsolved integer sequences worked out

[ Front page ] [ The edge ] [ Finished ] [ About ] [ Links ]


29.9 per cent

Held by Claude Opus 4.8. Set 2026-05-28, and has stood for 124 days. Written down here 2026-09-07.

The On-Line Encyclopedia of Integer Sequences has catalogued number patterns since 1964. Some of its entries are open: the sequence is listed, and the rule that produces it has never been worked out. Those open entries make a test with no answer key, which is a rare and valuable thing. The best model manages under a third of them. Like the Erdos problems this measures discovery rather than recall, but it is a good deal easier, which is why the figure is ten times higher. The two of them together bracket the question worth asking. Movement here is movement on the only sort of problem where a machine cannot quietly be scoring itself against something a person already published and it merely learned.

That description was written when this frontier was opened, on 2026-09-07, and is never rewritten. Where it and the number above it disagree, the number is the reading and the description is history.


What counts
The best score any model has recorded on the OEIS Open set, as published by Epoch AI - entries of the encyclopedia whose generating rule is not known.
What moves it
A model records a higher score on the same set.

Where to check it: https://epoch.ai/data/ai-benchmarking-dashboard

The program reads it from https://epoch.ai/data/benchmarks.csv, which is the same figure in a form a script can parse.


Last checked: 2026-09-27


Previous: Open problems in mathematics solved   Next: Research mathematics solved

[ Front page ] [ The edge ] [ Finished ] [ About ] [ Links ]