STaR: Bootstrapping Reasoning With Reasoning
This addresses the challenge of improving reasoning in language models for tasks like mathematics and commonsense question-answering, offering a more efficient alternative to large-scale data collection, though it is incremental as it builds on existing chain-of-thought methods.
The paper tackles the problem of inducing language models to generate step-by-step rationales for complex reasoning tasks without requiring massive rationale datasets, by proposing STaR, a bootstrapping technique that iteratively uses a small number of examples and a large dataset without rationales, resulting in significant performance improvements on multiple datasets and comparable performance to a 30x larger state-of-the-art model on CommonsenseQA.
Generating step-by-step "chain-of-thought" rationales improves language model performance on complex reasoning tasks like mathematics or commonsense question-answering. However, inducing language model rationale generation currently requires either constructing massive rationale datasets or sacrificing accuracy by using only few-shot inference. We propose a technique to iteratively leverage a small number of rationale examples and a large dataset without rationales, to bootstrap the ability to perform successively more complex reasoning. This technique, the "Self-Taught Reasoner" (STaR), relies on a simple loop: generate rationales to answer many questions, prompted with a few rationale examples; if the generated answers are wrong, try again to generate a rationale given the correct answer; fine-tune on all the rationales that ultimately yielded correct answers; repeat. We show that STaR significantly improves performance on multiple datasets compared to a model fine-tuned to directly predict final answers, and performs comparably to fine-tuning a 30$\times$ larger state-of-the-art language model on CommensenseQA. Thus, STaR lets a model improve itself by learning from its own generated reasoning.