AgentTune framework unifies training and evaluation for AI tool users
AgentTune: A Unified Training Lifecycle for Tool-Using LLM Agents
Machine Learning
Summary
Building AI agents that use tools well requires many steps, such as training, testing, and fixing them when they go wrong. The researchers created AgentTune, a software library that helps manage all these steps in one place using a special event-recording system. It supports different ways to train and evaluate these AI agents while allowing easy reuse and combination of parts. The library was shown to work on tasks like searching legal contracts and other specialized applications.
What this means in practice
- •For software engineering teams: Use AgentTune to build and optimize AI agents that integrate multiple tools and workflows reliably and with unified training and evaluation.
- •For legal technology developers: Develop domain-specific AI agents capable of tasks like contract search with a standardized pipeline for training and auditing behavior.
Authors
R E Zera Marveen Lyngkhoi, Abhivansh Gupta, Vidushee Vats, Ram Mohan Rao Kadiyala, Vinay Kumar Sankarapu, Pratinav Seth
Abstract
Tool-using agents increasingly require a common substrate for building, training, evaluating, distilling, and maintaining agentic systems, yet existing tooling typically addresses these stages in isolation. We present AgentTune, a source-available Python library that provides a unified framework for developing and optimizing tool-using agents. At its core, AgentTune introduces a two-tier EventLog representation that preserves agent trajectories while supporting consistent projections for training, evaluation, and auditing. Building on this substrate, the library provides a shared agent runtime, tool-calling reinforcement learning, agentic distillation, retrieval-augmented generation, decision workflows, evaluation utilities, and a designed path toward self-healing agents. These components are exposed through composable interfaces so that agent strategies, training methods, evaluation procedures, and recovery mechanisms can be combined without requiring separate representations or pipelines. We demonstrate the framework across agentic tasks including legal contract search and domain-specific applications, illustrating its use for trajectory-based optimization, tool-use evaluation, and cross-task agent development. AgentTune is released as a source-available library at https://github.com/Lexsi-Labs/AgentTune