Genko 原稿Back to edition

Beyond Selected Breakthroughs: Evaluating Language Models on Randomly Sampled Mathematical Conjectures

Benchmarks language models on randomly sampled conjectures, not famous ones.

Explainer · on the wire

The explainer for this story is still on the wire.

The claim, the mechanism, and the figures land here once generated.

Source · arXiv

arxiv.org/abs/2610.40382cs.AI · math.HO

More from this edition