Improving the ability of LLMs to act as world models that can help AI agents plan and execute goals.
World models generally refer to generative models of the transition functions in an MDP, able to aid an agent in selecting an optimal set of actions to complete a given task. LLMs, for all their internet scale knowledge, still struggle to be effective world models (see here) for embodied agents. And yet there is incredible potential in using large-scale data especially in the form of language to aid agents, we as people are able to read information via language and form generalizable world models internally that helps us achieve tasks. This project focuses on realizing such potential.
References
2023
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), 2023
@inproceedings{Nottingham2023Embodied,title={Do Embodied Agents Dream of Pixelated Sheep: Embodied Decision Making using Language Guided World Modelling},author={Nottingham, Kolby and Ammanabrolu, Prithviraj and Suhr, Alane and Choi, Yejin and Hajishirzi, Hannaneh and Singh, Sameer and Fox, Roy},booktitle={International Conference on Machine Learning (ICML)},url={https://arxiv.org/abs/2301.12050},year={2023},}
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), 2023
@inproceedings{lin2023swiftsage,title={SwiftSage: A Generative Agent with Fast and Slow Thinking for Complex Interactive Tasks},author={Lin, Bill Yuchen and Fu, Yicheng and Yang, Karina and Brahman, Faeze and Huang, Shiyu and Bhagavatula, Chandra and Ammanabrolu, Prithviraj and Choi, Yejin and Ren, Xiang},booktitle={Thirty-seventh Conference on Neural Information Processing Systems (NeurIPS)},year={2023},url={https://arxiv.org/abs/2305.17390},}
2022
ScienceWorld: Is your Agent Smarter than a 5th Grader?
Ruoyao Wang*, Peter Jansen*, Marc-Alexandre Côté, and Prithviraj Ammanabrolu
In Empirical Methods in Natural Language Processing (EMNLP), 2022
@inproceedings{wang2022scienceworld,title={ScienceWorld: Is your Agent Smarter than a 5th Grader?},author={Wang*, Ruoyao and Jansen*, Peter and Côté, Marc-Alexandre and Ammanabrolu, Prithviraj},booktitle={Empirical Methods in Natural Language Processing (EMNLP)},url={https://arxiv.org/abs/2203.07540},year={2022},}
2021
Modeling Worlds in Text
Prithviraj Ammanabrolu, and Mark Riedl
In The First Workshop on Commonsense Reasoning and Knowledge Bases (CSKB) at AKBC, 2021
@inproceedings{ammanabrolu2021modelingws,title={Modeling Worlds in Text},author={Ammanabrolu, Prithviraj and Riedl, Mark},booktitle={The First Workshop on Commonsense Reasoning and Knowledge Bases (CSKB) at AKBC},year={2021},url={https://openreview.net/forum?id=7FHnnENUG0},}
Modeling Worlds in Text
Prithviraj Ammanabrolu, and Mark Riedl
In Thirty-fifth Conference on Neural Information Processing Systems (NeurIPS) Datasets and Benchmarks Track (Round 1), 2021
@inproceedings{ammanabrolu2021modeling,title={Modeling Worlds in Text},author={Ammanabrolu, Prithviraj and Riedl, Mark},booktitle={Thirty-fifth Conference on Neural Information Processing Systems (NeurIPS) Datasets and Benchmarks Track (Round 1)},year={2021},url={https://openreview.net/forum?id=7FHnnENUG0},}
Learning Knowledge Graph-based World Models of Textual Environments
Prithviraj Ammanabrolu, and Mark Riedl
In Thirty-fifth Conference on Neural Information Processing Systems (NeurIPS), 2021
@inproceedings{ammanabrolu2021learning,title={Learning Knowledge Graph-based World Models of Textual Environments},author={Ammanabrolu, Prithviraj and Riedl, Mark},booktitle={Thirty-fifth Conference on Neural Information Processing Systems (NeurIPS)},year={2021},url={https://arxiv.org/abs/2106.09608},}