![[2602.14295] Machine Learning as a Tool (MLAT): A Framework for Integrating Statistical ML Models as Callable Tools within LLM Agent Workflows](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.14295] Machine Learning as a Tool (MLAT): A Framework for Integrating Statistical ML Models as Callable Tools within LLM Agent Workflows
Summary
The paper introduces Machine Learning as a Tool (MLAT), a framework for integrating statistical ML models as callable tools within LLM workflows, enhancing contextual reasoning and efficiency.
Why It Matters
MLAT represents a significant advancement in how machine learning models can be utilized within large language models (LLMs). By treating ML models as first-class tools, it allows for more dynamic and context-aware applications, which can improve decision-making processes in various domains, particularly in generating proposals and estimates.
Key Takeaways
- MLAT allows LLMs to invoke statistical models as tools based on context.
- The framework enhances efficiency in generating proposals, reducing time from hours to minutes.
- PitchCraft demonstrates MLAT's application in real-world scenarios, achieving high predictive accuracy.
- MLAT can generalize to various domains requiring quantitative estimation.
- The framework addresses challenges of data scarcity in training ML models.
Computer Science > Machine Learning arXiv:2602.14295 (cs) [Submitted on 15 Feb 2026] Title:Machine Learning as a Tool (MLAT): A Framework for Integrating Statistical ML Models as Callable Tools within LLM Agent Workflows Authors:Edwin Chen, Zulekha Bibi View a PDF of the paper titled Machine Learning as a Tool (MLAT): A Framework for Integrating Statistical ML Models as Callable Tools within LLM Agent Workflows, by Edwin Chen and Zulekha Bibi View PDF HTML (experimental) Abstract:We introduce Machine Learning as a Tool (MLAT), a design pattern in which pre-trained statistical machine learning models are exposed as callable tools within large language model (LLM) agent workflows. This allows an orchestrating agent to invoke quantitative predictions when needed and reason about their outputs in context. Unlike conventional pipelines that treat ML inference as a static preprocessing step, MLAT positions the model as a first-class tool alongside web search, database queries, and APIs, enabling the LLM to decide when and how to use it based on conversational context. To validate MLAT, we present PitchCraft, a pilot production system that converts discovery call recordings into professional proposals with ML-predicted pricing. The system uses two agents: a Research Agent that gathers prospect intelligence via parallel tool calls, and a Draft Agent that invokes an XGBoost pricing model as a tool call and generates a complete proposal through structured outputs. The pricing model,...
