Glossary · Term

OlympiadBench

← all terms

Definition

Plain language

A benchmark of olympiad-level math problems used to evaluate advanced reasoning models.

As stated in the literature

A multi-discipline olympiad-style benchmark covering mathematics and physics at advanced competition difficulty, used in reasoning-model evaluation suites alongside MATH and AIME.

Why it matters: It pushes evaluation beyond standard math sets into multi-step problems where proof structure and physical reasoning actually matter.

For example, a problem might ask the model to derive the trajectory of a charged particle in crossed electric and magnetic fields.

Heard on the show

“Across six math benchmarks — MATH-500, GSM8K, AMC, AIME, Minerva, OlympiadBench — and across model families and sizes, ReasonMaxxer matches or exceeds full RL pipelines on the majority of comparisons.”
Episode 026 — What RL Actually Does to Language Models, at the Token Level

Mentioned in 1 episode

  1. 026
    What RL Actually Does to Language Models, at the Token Level

Related terms