Auto-SLURP: A Benchmark Dataset for Evaluating Multi-Agent Frameworks in Smart Personal Assistant
Lei Shen, Xiaoyu Shen
TL;DR
Auto-SLURP addresses the lack of end-to-end benchmarks for evaluating LLM-based multi-agent personal assistants by extending the SLURP dataset with relabeled slots and simulated servers to test language understanding, orchestration, and backend execution. It compares multiple GPT-4–driven frameworks under a standardized workflow, revealing that end-to-end reliability remains challenging and that targeted finetuning of the intent component yields substantial gains. Spanning 23 domains with integrated external services, the benchmark provides a realistic testbed for assessing decision making, tool use, and coordination in intelligent assistants. The findings highlight the need for improved orchestration policies and prompting strategies to realize dependable AI-powered personal assistants in practice.
Abstract
In recent years, multi-agent frameworks powered by large language models (LLMs) have advanced rapidly. Despite this progress, there is still a notable absence of benchmark datasets specifically tailored to evaluate their performance. To bridge this gap, we introduce Auto-SLURP, a benchmark dataset aimed at evaluating LLM-based multi-agent frameworks in the context of intelligent personal assistants. Auto-SLURP extends the original SLURP dataset -- initially developed for natural language understanding tasks -- by relabeling the data and integrating simulated servers and external services. This enhancement enables a comprehensive end-to-end evaluation pipeline, covering language understanding, task execution, and response generation. Our experiments demonstrate that Auto-SLURP presents a significant challenge for current state-of-the-art frameworks, highlighting that truly reliable and intelligent multi-agent personal assistants remain a work in progress. The dataset and related code are available at https://github.com/lorashen/Auto-SLURP/.