AoT
Atom of Thoughts (AoT) is a framework used to enhance the inference performance of large language models.
0
- Support multiple reasoning scenarios, including mathematics, multiple-choice questions, and multi hop question answering
- Can be combined with existing testing methods as a plugin to improve performance
- By representing atomic states, improve the efficiency of computing resource utilization
- Supports multiple datasets and models, such as gpt-4o-mini
- Provide quick startup scripts for users to conduct experiments and tests conveniently
- Support generating compressed datasets for integration with other frameworks
- Open source code facilitates secondary development for researchers and developers
Product Details
Atom of Thoughts (AoT) is a novel reasoning framework that transforms the reasoning process into a Markov process by representing solutions as combinations of atomic problems. This framework significantly improves the performance of large language models in inference tasks through decomposition and contraction mechanisms, while reducing the waste of computing resources. AoT can not only serve as an independent inference method, but also as a plugin for extending existing testing methods, flexibly combining the advantages of different methods. This framework is open source and based on Python implementation, suitable for researchers and developers to conduct experiments and applications in the fields of natural language processing and large language modeling.
