BigHugger
GH Repository · xingyaoww

code-act

Official Repo for ICML 2024 paper "Executable Code Actions Elicit Better LLM Agents" by Xingyao Wang, Yangyi Chen, Lifan Yuan, Yizhe Zhang, Yunzhu Li, Hao Peng, Heng Ji.

stars
1,703
30-day movement
starts with the next reading
Related entries
60
Connections
1
python/condapythonPythonagent-frameworkllm-frameworkllm-finetuningllm-agentllm

This is the official code repository for the ICML 2024 paper "Executable Code Actions Elicit Better LLM Agents", which proposes having LLM agents take actions by writing executable code. It is written in Python, installable via conda, and released under an MIT licence.

You want the reference implementation and code accompanying a published research result on code-action-based LLM agents.

Use it to

  • Reproduce the ICML 2024 paper's experiments
  • Build agents whose actions are executable code
  • Fine-tune LLMs for code-action agent behaviour
  • Extend the framework for your own agent research

For Researchers and developers building code-action LLM agents

Role
agent-framework
Language
Python
Licence
MIT
Forks
146
Open issues
12
Last push
2024-05-23
topicsllmllm-agentllm-finetuningagent-frameworkresearch-codepython