Search papers, labs, and topics across Lattice.
This paper introduces Agentic-imodels, an autoresearch loop that evolves interpretable machine learning models specifically for agentic data science (ADS) systems. The key innovation is a novel LLM-based interpretability metric that assesses how well an LLM can "simulate" a model's behavior based on its string representation. Experiments show that evolved models improve both predictive performance and agent-facing interpretability, leading to significant gains (up to 73%) on the BLADE benchmark for downstream ADS tasks.
Forget human-readable models: Agentic-imodels evolves ML models that are optimized for LLM interpretability, boosting agentic data science performance by up to 73%.
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%
Microsoft Research