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

2026-04-17 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

EVIL (Evolving Interpretable algorithms with LLMs) is a novel approach that employs LLM-guided evolutionary search to discover simple, interpretable Python/NumPy algorithms for dynamical systems inference. Instead of training neural networks on extensive datasets, EVIL evolves pure programs capable of zero-shot, in-context inference across various datasets. The method was applied 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 demonstrated generalization across all evaluation datasets without requiring per-dataset training. These discovered algorithms often achieve competitive or superior performance compared to state-of-the-art deep learning models, while being orders of magnitude faster and fully interpretable.

Key takeaway

For AI Engineers and Research Scientists developing inference models for dynamical systems, EVIL demonstrates a compelling alternative to deep learning. You should consider exploring LLM-guided program evolution to develop highly interpretable, fast, and generalizable algorithms, especially when zero-shot performance across diverse datasets is a critical requirement. This approach could significantly reduce training overhead and improve model transparency.

Key insights

LLM-guided evolutionary search can discover interpretable, zero-shot algorithms for dynamical systems inference.

Principles

• Interpretability enhances algorithm utility.
• Zero-shot generalization is achievable via program evolution.

Method

EVIL uses LLM-guided evolutionary search to generate and refine pure Python/NumPy programs, enabling zero-shot, in-context inference for dynamical systems tasks without traditional neural network training.

In practice

• Apply LLM-guided evolution for algorithm discovery.
• Prioritize interpretable solutions for complex systems.

Topics

• EVIL
• LLM-guided Evolutionary Search
• Interpretable Algorithms
• Zero-Shot Inference
• Dynamical Systems Inference

Best for: AI Engineer, Research Scientist, AI Scientist, Machine Learning Engineer

Related on AIssential

• EVIL: Evolving Interpretable Algorithms for Zero-Shot Inference on Event Sequences and Time Series with LLMs — LLM-guided evolution can discover compact, interpretable algorithms for zero-shot dynamical systems inference, outperforming complex neural networks.
• Capture the Flags: Family-Based Evaluation of Agentic LLMs via Semantics-Preserving Transformations — CTF challenge families and Evolve-CTF enable robust evaluation of agentic LLMs against code transformations.
• Mutation Without Variation: Convergence Dynamics in LLM-Driven Program Evolution — LLM-driven program mutation inherently converges to limited structural forms, hindering open-ended exploration.
• LLM-Guided Evolution for Medical Decision Pipelines — LLM-guided evolution optimizes medical decision pipelines at inference-time, outperforming manual baselines through interpretable program-level changes.
• Data Machina #261 — Generative AI and LLMs are being adapted to achieve state-of-the-art performance in diverse time-series tasks.
• RECAP: Regression Evaluation for Continual Adaptation of Prompts — Existing prompt adaptation methods are structurally inadequate for proactive, real-time constraint evolution in agentic systems.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.