Agentic-imodels: Evolving agentic interpretability tools via autoresearch

TL;DR

This paper introduces Agentic-imodels, an autoresearch system that evolves interpretable data science tools optimized for agentic understanding, improving predictive performance and interpretability for autonomous data analysis.

Contribution

It develops a library of regressors optimized for both accuracy and a novel LLM-based interpretability metric, advancing agent-centric interpretability in data science tools.

Findings

Evolved models improve predictive accuracy and interpretability.

Models generalize to new datasets and interpretability tests.

Enhanced models boost downstream ADS performance by up to 73%.

Abstract

Agentic data science (ADS) systems are rapidly improving their capability to autonomously analyze, fit, and interpret data, potentially moving towards a future where agents conduct the vast majority of data-science work. However, current ADS systems use statistical tools designed to be interpretable by humans, rather than interpretable by agents. To address this, we introduce Agentic-imodels, an agentic autoresearch loop that evolves data-science tools designed to be interpretable by agents. Specifically, it develops a library of scikit-learn-compatible regressors for tabular data that are optimized for both predictive performance and a novel LLM-based interpretability metric. The metric measures a suite of LLM-graded tests that probe whether a fitted model's string representation is "simulatable" by an LLM, i.e. whether the LLM can answer questions about the model's behavior by reading its string output alone. We find that the evolved models jointly improve predictive performance and agent-facing interpretability, generalizing to new datasets and new interpretability tests. Furthermore, these evolved models improve downstream end-to-end ADS, increasing performance for Copilot CLI, Claude Code, and Codex on the BLADE benchmark by up to 73%