Skip to content

linyuhongg/LLM-based-Optimization-of-Compound-AI-Systems

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 

Repository files navigation

LLM-based Optimization of Compound AI Systems

alt text

This repository is a collection of influential papers and resources related to LLM-based optimization of compound AI systems as explored in the survey paper titled “LLM-based Optimization of Compound AI Systems: A Survey”.

For more information please refer to our survey paper or twitter thread.

📚 Papers

alt text

1. Workflow: Design of a Compound AI System

1.1 LLM with Tools

1.2 LLM with Code Interpreter

1.3 Retrieval Augmented Generation

1.4 Search with LLM Calls

  • Title: Language Agent Tree Search Unifies Reasoning, Acting, and Planning in Language Models, ICML, 2024.
    Authors: Andy Zhou, Kai Yan, Michal Shlapentokh-Rothman, Haohan Wang, Yu-Xiong Wang TL;DR: MCTS over action space with multiple LLM calls where we have a generator and a evaluator.

  • Title: Tree of Thoughts: Deliberate Problem Solving with Large Language Models, Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023, 2023.
    Authors: Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, Karthik Narasimhan TL;DR: BFS/DFS with multiple LLM calls where we have a generator and a evaluator.

  • Title: Reasoning with Language Model is Planning with World Model, EMNLP, 2023.
    Authors: Shibo Hao, Yi Gu, Haodi Ma, Joshua Jiahua Hong, Zhen Wang, Daisy Zhe Wang, Zhiting Hu TL;DR: MCTS over reasoning space with multiple LLM calls where we have a generator and a evaluator.

  • Title: Self-Consistency Improves Chain of Thought Reasoning in Language Models, ICLR, 2023.
    Authors: Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc V. Le, Ed H. Chi, Sharan Narang, Aakanksha Chowdhery, Denny Zhou TL;DR: CoT@k.

  • Title: Reflexion: language agents with verbal reinforcement learning, Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023, 2023.
    Authors: Noah Shinn, Federico Cassano, Ashwin Gopinath, Karthik Narasimhan, Shunyu Yao TL;DR: Requires an environment that allows to retry actions. It uses a previously failed action and uses an LLM to reflect on it to come up with a better action.

  • Title: Self-Refine: Iterative Refinement with Self-Feedback, Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023, 2023.
    Authors: Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, Shashank Gupta, Bodhisattwa Prasad Majumder, Katherine Hermann, Sean Welleck, Amir Yazdanbakhsh, Peter Clark TL;DR: Generate an initial output using an LLM; then, the same LLM provides feedback for its output and uses it to refine itself, iteratively.

  • Title: MAgICoRe: Multi-Agent, Iterative, Coarse-to-Fine Refinement for Reasoning, Arxiv, 2024.
    Authors: Justin Chih-Yao Chen and Archiki Prasad and Swarnadeep Saha and Elias Stengel-Eskin and Mohit Bansal TL;DR: Generate an initial output using an LLM; then, the same LLM provides feedback for its output and uses it to refine itself, iteratively.

2. Optimization

🛠️ How to Contribute

Contributions to this repository are welcome! If you would like to add a new paper or resource, please follow these steps:

1.	Fork the repository.
2.	Create a new branch (git checkout -b feature-paper-contribution).
3.	Add your paper or resource to the papers/ directory.
4.	Commit your changes (git commit -m 'Added new paper: [Paper Title]').
5.	Push to the branch (git push origin feature-paper-contribution).
6.	Open a Pull Request and provide a brief description of your contribution.

📬 Contact

For any questions or feedback, feel free to reach out to me at [email protected]

📝 How to Cite the Survey

If you find this repository useful in your research, please consider citing the survey paper:

@misc{lin2024llmbasedoptimizationcompoundai,
      title={LLM-based Optimization of Compound AI Systems: A Survey}, 
      author={Matthieu Lin and Jenny Sheng and Andrew Zhao and Shenzhi Wang and Yang Yue and Yiran Wu and Huan Liu and Jun Liu and Gao Huang and Yong-Jin Liu},
      year={2024},
      eprint={2410.16392},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2410.16392}, 
}

About

LLM-based Optimization of Compound AI Systems: A Survey

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published