Datasets:
Tasks:
Text Generation
Languages:
English
Size:
1K - 10K
ArXiv:
Tags:
Formal theorem proving
Lean 4
Stochastic processes
Mathematical reasoning
Natural language processing
Autoformalization
License:
Commit History
-
Benchmark Release V1.0
Verified signatureLeading benchmarks for formal theorem proving with large language models are small collections drawn from competition math, such as the IMO and Putnam, that poorly represent field-specific applications. We introduce StochBench, a Lean 4 benchmark of 450 graduate stochastic-processes problems at varying abstraction levels, each paired with its natural-language source. Addressing a field underrepresented in Mathlib, it covers finite and countable Markov chains, renewal processes, random walks, martingales, stopping times, queues, Brownian motion, stochastic calculus, weak convergence, and Poisson and continuous-time Markov processes. Our Opus 4.8-based agent achieves a 34.9% proof rate (157/450) under a 15-minute per-problem limit. StochBench better represents domain-specific applied mathematics while remaining challenging for advanced provers. 335ab0f