🪄 Lumos: Language Agents with Unified Formats, Modular Design, and Open-Source LLMs

<p align="center"> 🌐<a href="https://allenai.github.io/lumos">[Website]</a>   📝<a href="">[Paper]</a>   🤗<a href="https://huggingface.co/datasets?sort=trending&search=ai2lumos">[Data]</a>   🤗<a href="https://huggingface.co/models?sort=trending&search=ai2lumos">[Model]</a>   </p>

We introduce 🪄Lumos, Language Agents with Unified Formats, Modular Design, and Open-Source LLMs. Lumos unifies a suite of complex interactive tasks and achieves competitive performance with GPT-4/3.5-based and larger open-source agents.

Lumos has following features:

• 🧩 Modular Architecture:
  • Lumos consists of planning, grounding, and execution modules built based on LLAMA-2-7B.
• 🌍 Diverse Training Data:
  • Lumos is trained with ~40K high-quality annotations from ground-truth reasoning steps in existing benchmarks with GPT-4.
• 🚀 Competitive Performance:
  • 🚀 Lumos outperforms GPT-4/3.5-based agents on complex QA and web agent tasks, and larger open agents on maths tasks.
  • 🚀 Lumos performs better than open agent baseline formulations including chain-of-thoughts and unmodularized training.
  • 🚀 Lumos surpasses larger open LLM agents and domain-specific agents on an unseen task, WebShop.

Model Overview

lumos_web_agent_ground_iterative is a grounding module checkpoint finetuned on web agent task in Lumos-Iterative (Lumos-I) formulation.

The training annotation is shown below:

Citation

If you find this work is relevant with your research, please feel free to cite our work!

@article{yin2023lumos,
  title={Lumos: Towards Language Agents that are Unified, Modular, and Open Source},
  author={Yin, Da and Brahman, Faeze and Ravichander, Abhilasha and Chandu, Khyathi and Chang, Kai-Wei and Choi, Yejin and Lin, Bill Yuchen},
  year={2023}
}