MAG-V: A Multi-Agent Framework for Synthetic Data Generation and Verification

TL;DR

MAG-V is a multi-agent framework that generates synthetic customer queries and verifies agent response trajectories, improving agent performance and ensuring reliability without extensive real-world data collection.

Contribution

The paper introduces MAG-V, a novel multi-agent system for synthetic data generation and trajectory verification, enhancing agent testing and alignment with minimal real data.

Findings

Synthetic data improves agent performance on real queries.

Trajectory verification outperforms GPT-4o judge by 11% accuracy.

Method matches GPT-4 judge performance on constructed dataset.

Abstract

Extending the capabilities of Large Language Models (LLMs) with functions or tools for environment interaction has led to the emergence of the agent paradigm. In industry, training an LLM is not always feasible because of the scarcity of domain data, legal holds on proprietary customer data, rapidly changing business requirements, and the need to prototype new assistants. Agents provide an elegant solution to the above by relying on the zero-shot reasoning abilities of the underlying LLM and utilizing tools to explore and reason over customer data and respond to user requests. However, there are two concerns here: (I) acquiring large scale customer queries for agent testing is time-consuming, and (II) high reliance on the tool call sequence (or trajectory) followed by the agent to respond to user queries may lead to unexpected or incorrect behavior. To address this, we propose MAG-V, a multi-agent framework to first generate a dataset of questions that mimic customer queries; and second, reverse-engineer alternate questions from the responses for trajectory verification. Initial results indicate that our synthetic data can improve agent performance on actual customer queries. Furthermore, our trajectory verification methodology, inspired by distant supervision and using traditional machine learning (ML) models, outperforms a GPT-4o judge baseline by 11% accuracy and matches the performance of a GPT-4 judge on our constructed dataset. Overall, our approach is a step towards unifying diverse task agents into a cohesive framework for achieving an aligned objective.