Husky: A Unified, Open-Source Language Agent for Multi-Step Reasoning
This provides a holistic, open-source solution for researchers and practitioners needing versatile language agents for complex reasoning tasks, though it builds incrementally on existing agent frameworks.
The authors tackled the problem of language agents being proprietary or task-specific by introducing Husky, an open-source agent that uses a unified action space for multi-step reasoning across diverse tasks, achieving state-of-the-art performance on 14 datasets and matching or exceeding GPT-4 on new benchmarks despite using smaller 7B models.
Language agents perform complex tasks by using tools to execute each step precisely. However, most existing agents are based on proprietary models or designed to target specific tasks, such as mathematics or multi-hop question answering. We introduce Husky, a holistic, open-source language agent that learns to reason over a unified action space to address a diverse set of complex tasks involving numerical, tabular, and knowledge-based reasoning. Husky iterates between two stages: 1) generating the next action to take towards solving a given task and 2) executing the action using expert models and updating the current solution state. We identify a thorough ontology of actions for addressing complex tasks and curate high-quality data to train expert models for executing these actions. Our experiments show that Husky outperforms prior language agents across 14 evaluation datasets. Moreover, we introduce HuskyQA, a new evaluation set which stress tests language agents for mixed-tool reasoning, with a focus on retrieving missing knowledge and performing numerical reasoning. Despite using 7B models, Husky matches or even exceeds frontier LMs such as GPT-4 on these tasks, showcasing the efficacy of our holistic approach in addressing complex reasoning problems. Our code and models are available at https://github.com/agent-husky/Husky-v1.