ReflectEvo: Improving Meta Introspection of Small LLMs by Learning Self-Reflection

TL;DR

ReflectEvo introduces a self-reflection learning pipeline for small language models, significantly improving their reasoning abilities through iterative self-generated reflection and a large-scale dataset, surpassing some open-source models.

Contribution

The paper presents a novel reflection learning pipeline and a large-scale reflection dataset, demonstrating substantial reasoning improvements in small language models without human annotations.

Findings

Llama-3's reasoning accuracy improved from 52.4% to 71.2%.

Mistral's reasoning accuracy improved from 44.4% to 71.1%.

ReflectEvo can rival or surpass open-source models on BIG-bench.

Abstract

We present a novel pipeline, ReflectEvo, to demonstrate that small language models (SLMs) can enhance meta introspection through reflection learning. This process iteratively generates self-reflection for self-training, fostering a continuous and self-evolving process. Leveraging this pipeline, we construct ReflectEvo-460k, a large-scale, comprehensive, self-generated reflection dataset with broadened instructions and diverse multi-domain tasks. Building upon this dataset, we demonstrate the effectiveness of reflection learning to improve SLMs' reasoning abilities using SFT and DPO with remarkable performance, substantially boosting Llama-3 from 52.4% to 71.2% and Mistral from 44.4% to 71.1%. It validates that ReflectEvo can rival or even surpass the reasoning capability of the three prominent open-sourced models on BIG-bench without distillation from superior models or fine-grained human annotation. We further conduct a deeper analysis of the high quality of self-generated reflections and their impact on error localization and correction. Our work highlights the potential of continuously enhancing the reasoning performance of SLMs through iterative reflection learning in the long run.