Reasoning & Test-Time Compute · 2025

rStar-Math: Small LLMs Can Master Math Reasoning with Self-Evolved Deep Thinking

Xinyu Guan, Mao Yang

Lets small models rival o1 on MATH via MCTS test-time search guided by a process preference model, with a self-evolution loop that trains policy and reward model from scratch without distillation.

Editorial record

Plain-language summary

Lets small models rival o1 on MATH via MCTS test-time search guided by a process preference model, with a self-evolution loop that trains policy and reward model from scratch without distillation.

Knowledge graph

Relationships

Descendants

Source record

Provenance

Record ID
P-640
Record created
2026-07-13
Last reviewed
2026-07-14
Record version
2

Citation caveat: Citation metadata is approximate and marked unverified in the source dataset.