EVIL: Evolving Interpretable Algorithms for Zero-Shot Inference on Event Sequences and Time Series with LLMs

2026-04-17

AI Summary

This work proposes a training-free, zero-shot inference method for modeling dynamical systems from event sequences and time series. By leveraging large language model (LLM)-guided evolutionary search, the approach automatically synthesizes concise, interpretable Python/NumPy programs that enable unified cross-dataset inference. For the first time, it demonstrates that LLM-guided program evolution can discover compact algorithms generalizing across multiple classes of dynamical systems. Using a single evolutionary strategy, the method achieves cross-dataset generalization on three distinct tasks, matching or surpassing state-of-the-art deep learning models in performance while offering several orders of magnitude faster inference and full interpretability.

Technology Category

• Machine Learning: Evolutionary Learning
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Search and Optimization: Evolutionary Computation

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

We introduce EVIL (\textbf{EV}olving \textbf{I}nterpretable algorithms with \textbf{L}LMs), an approach that uses LLM-guided evolutionary search to discover simple, interpretable algorithms for dynamical systems inference. Rather than training neural networks on large datasets, EVIL evolves pure Python/NumPy programs that perform zero-shot, in-context inference across datasets. We apply EVIL to three distinct tasks: next-event prediction in temporal point processes, rate matrix estimation for Markov jump processes, and time series imputation. In each case, a single evolved algorithm generalizes across all evaluation datasets without per-dataset training (analogous to an amortized inference model). To the best of our knowledge, this is the first work to show that LLM-guided program evolution can discover a single compact inference function for these dynamical-systems problems. Across the three domains, the discovered algorithms are often competitive with, and even outperform, state-of-the-art deep learning models while being orders of magnitudes faster, and remaining fully interpretable.

Problem

• zero-shot inference
• interpretable algorithms
• event sequences
• time series
• dynamical systems

Innovation

• LLM-guided evolution
• interpretable algorithms
• zero-shot inference
• program synthesis
• dynamical systems