Self-Reflection in LLM Agents: Effects on Problem-Solving Performance
What it is
This repository contains the code and data for a study investigating how self-reflection improves the problem-solving performance of Large Language Models (LLMs). It provides scripts to run baseline agents, generate self-reflections on mistakes, and re-evaluate performance using that guidance across nine popular LLMs.
The problem it solves
LLMs often make errors in problem-solving; this project addresses the need to understand how and if self-correction via reflection can significantly improve their accuracy.
What you could use it for
- Reproducing the study on self-reflection in LLM agents
- Analyzing the impact of different reflection types on multiple-choice question accuracy
- Accessing a dataset of LLM dialogues and reflections for further research
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.