Tool Use

Teach AI agents to more effectively use existing tools and APIs.

Why teach your LLM to add numbers when you can teach it to use a calculator instead? This project explores techniques for augmenting interactive AI agents with information from symbolic modules, much like humans use tools like calculators and GPS systems to assist with arithmetic and navigation.

References

2023

  1. Behavior Cloned Transformers are Neurosymbolic Reasoners
    Ruoyao Wang*, Peter Jansen*, Marc-Alexandre Côté, and Prithviraj Ammanabrolu
    In European Chapter of the Association for Computational Linguistics (EACL), Apr 2023
  2. Do Embodied Agents Dream of Pixelated Sheep: Embodied Decision Making using Language Guided World Modelling
    Kolby Nottingham, Prithviraj Ammanabrolu, Alane Suhr, Yejin Choi, Hannaneh Hajishirzi, Sameer Singh, and Roy Fox
    In International Conference on Machine Learning (ICML), Apr 2023
  3. SwiftSage: A Generative Agent with Fast and Slow Thinking for Complex Interactive Tasks
    Bill Yuchen Lin, Yicheng Fu, Karina Yang, Faeze Brahman, Shiyu Huang, Chandra Bhagavatula, Prithviraj Ammanabrolu, Yejin Choi, and Xiang Ren
    In Thirty-seventh Conference on Neural Information Processing Systems (NeurIPS), Apr 2023