API-BLEND: A Comprehensive Corpora for Training and Benchmarking API LLMs
This addresses the problem of data scarcity for researchers and developers working on tool-augmented LLMs, but it is incremental as it builds on existing datasets rather than creating new paradigms.
The paper tackles the need for data to train and benchmark LLMs that use APIs by introducing API-BLEND, a large corpora curated from existing datasets to simulate real-world API tasks, and demonstrates its utility for both training and benchmarking.
There is a growing need for Large Language Models (LLMs) to effectively use tools and external Application Programming Interfaces (APIs) to plan and complete tasks. As such, there is tremendous interest in methods that can acquire sufficient quantities of train and test data that involve calls to tools / APIs. Two lines of research have emerged as the predominant strategies for addressing this challenge. The first has focused on synthetic data generation techniques, while the second has involved curating task-adjacent datasets which can be transformed into API / Tool-based tasks. In this paper, we focus on the task of identifying, curating, and transforming existing datasets and, in turn, introduce API-BLEND, a large corpora for training and systematic testing of tool-augmented LLMs. The datasets mimic real-world scenarios involving API-tasks such as API / tool detection, slot filling, and sequencing of the detected APIs. We demonstrate the utility of the API-BLEND dataset for both training and benchmarking purposes.