Skip to main content
PaperLLMsv1.0

STaR: Bootstrapping Reasoning With Reasoning

by Stanford University / Google Brain · free · Last verified 2026-03-17

Introduced STaR (Self-Taught Reasoner), an iterative bootstrapping algorithm that enables a model to improve its reasoning capabilities using its own generated rationales. STaR alternates between generating rationales, filtering correct ones, and fine-tuning, achieving near-GPT-3-level reasoning from a much smaller model.

https://arxiv.org/abs/2203.14465
B
BAbove Average
Adoption: B+Quality: A+Freshness: B+Citations: B+Engagement: F

Specifications

License
Open Access
Pricing
free
Capabilities
self-improvement, rationale-generation, bootstrapping, math-reasoning
Integrations
Use Cases
iterative-reasoning-training, self-improvement, small-model-enhancement
API Available
No
Tags
star, self-taught-reasoner, bootstrapping, reasoning, rationale-generation
Added
2026-03-17
Completeness
100%

Index Score

67.5
Adoption
75
Quality
90
Freshness
72
Citations
78
Engagement
0

Put AI to work for your business

Deploy this paper alongside autonomous AaaS agents that handle tasks end-to-end — no babysitting required.

Explore the full AI ecosystem on Agents as a Service