oeis

LANTERN is a fast pipeline that uses a classifier over pretrained language model activations to rank candidate mathematical relations, then filters, generates, and verifies hypotheses. Applied to OEIS, it found 62 verified relations, 9 informative ones, and 4 entirely novel mathematical connections, completing the whole process in under 8 hours.

This paper analyzes integer-sequence benchmarks for language models using minimum description length, revealing that these benchmarks often measure memorisation rather than inductive reasoning, and introduces a new difficulty measure.

Explores the OEIS meta sequence A051070, where each term is the n-th term of sequence A_n, highlighting quirky entries involving New York subway stops and the paradoxical self-reference of sequence A102288.