Keyword Search Matches 8B Embedding Model in Agent Tool Selection

Written by

in

In tests involving 200 ToolRet questions and up to 3,794 tools, BM25 ran on CPU only, was about 40 times faster than BGE-large, and matched the performance of an 8B embedding model in agent tool selection. When using the full tool documentation, BM25 was the best-performing configuration in the tests.

Source:Towards Data Science · Read Original

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *