Code for paper "CRAFT: Customizing LLMs by Creating and Retrieving from Specialized Toolsets"
Previous approaches directly solve the given problem by generating code solutions, which may contain errors. CRAFT first creates a toolset that contains diverse, reusable, and correct tools that are executable code snippets. During inference, CRAFT employs a multi-view matching approach, incorporating information about the target problem, API names, and docstrings, to identify and utilize relevant tools, enhancing its problem-solving capabilities.
In toolset creation, we apply an iterative problem-sampling strategy based on similarity for diversity, followed by generating code solutions using GPT-4. To ensure the reusability of the created tools, we abstract the specific solutions into high-level tools that can tackle various kinds of problems by instructing GPT-4. To ensure the tools' correctness, we evaluate the tools on the original problems and discard those outputting incorrect answers. Finally, we deduplicate the tools to reduce redundancy, and finally obtain a toolset.
In inference, we apply a multi-view matching algorithm regarding the target problem, function name, and docstring between those in the toolset to retrieve related tools.
There are two folders in this repository, vqa and tab_and_math, representing three tasks considered in the paper: VQA, tabular processing, and mathematics reasoning. VQA is implemented based on ViperGPT, and the other two are built on CREATOR.
For VQA, cd vqa and then follow the guidance in ViperGPT to install the environment and download necessary models. Run construct_toolset.py to create VQA-specific tool functions and run run_exp.py to solve downstream problems with retrieval.
For the other two tasks, cd tab_and_math and install the environment by running pip install -r requirements.txt. The code for toolset construction is in construct_toolset.py, and the code for running experiments on downstream datasets is in generate_with_retrieval.py.
Please kindly cite our paper if you find our method or code useful:
@article{yuan2023craft,
title={CRAFT: Customizing LLMs by Creating and Retrieving from Specialized Toolsets},
author={Yuan, Lifan and Chen, Yangyi and Wang, Xingyao and Fung, Yi R. and Peng, Hao and Ji, Heng},
journal={arXiv preprint arXiv:2309.17428},
year={2023}
}