DaYin's picture
Update README.md
58a4d58 verified
metadata
license: apache-2.0
datasets:
  - ai2lumos/lumos_maths_ground_iterative
language:
  - en
tags:
  - language-agent
  - maths
  - reasoning
  - grounding

πŸͺ„ Agent Lumos: Unified and Modular Training for Open-Source Language Agents

🌐[Website]   πŸ“[Paper]   πŸ€—[Data]   πŸ€—[Model]   πŸ€—[Demo]  

We introduce πŸͺ„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/13B and off-the-shelf APIs.
    • πŸ€— Lumos utilizes a unified data format that encompasses multiple task types, thereby enabling the developed agent framework to conveniently support a range of interactive tasks.
  • 🌍 Diverse Training Data:
    • 🌍 Lumos is trained with ~56K diverse high-quality subgoal/action annotations from ground-truth reasoning steps in existing benchmarks with GPT-4.
    • βš’οΈ Lumos data can be instrumental for future research in developing open-source agents for complex interactive tasks.
  • πŸš€ Competitive Performance:
    • πŸš€ Lumos is comparable or even beats GPT-series agents on web/complex QA tasks Mind2Web and HotpotQA, and larger open agents on math and multimodal tasks.
    • πŸš€ Lumos exceeds contemporaneous agents that have been fine-tuned with in-domain HotpotQA, Mind2Web and ScienceQA annotations, such as FiReAct, AgentLM, and AutoAct.
    • πŸš€ Lumos performs better than open agent baseline formulations including chain-of-thoughts and integrated training.
    • πŸš€ Lumos surpasses larger open LLM agents and domain-specific agents on unseen tasks, WebShop and InterCode_SQL.

Model Overview

lumos_maths_ground_iterative is a grounding module checkpoint finetuned on maths task in Lumos-Iterative (Lumos-I) formulation.

The training annotation is shown below:

Training Data Number
lumos_maths_ground_iterative 19778

Citation

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

@article{yin2023lumos,
  title={Agent Lumos: Unified and Modular Training for Open-Source Language Agents},
  author={Yin, Da and Brahman, Faeze and Ravichander, Abhilasha and Chandu, Khyathi and Chang, Kai-Wei and Choi, Yejin and Lin, Bill Yuchen},
  journal={arXiv preprint arXiv:2311.05657},
  year={2023}
}