Towards Data Science is a community-powered publication that showcases work in data science, machine learning and artificial intelligence. Every day newcomers, seasoned researchers and industry practitioners publish tutorials, research notes and real-world case studies that help the field move forward.

Towards Data Science

Recursive Language Models (RLMs) address context window limitations and context rot by treating prompts as variables in a Python REPL environment rather than feeding everything into a single prompt. Instead of context summarization, RLMs let the LLM write code to programmatically inspect data and recursively invoke itself on smaller fragments. DSPy 3.1.2+ provides built-in RLM support. A practical demonstration shows analyzing 400K tokens of articles by having the model parse structure, filter content, and use batched recursive sub-calls to extract insights—effectively processing contexts up to 100x larger than the native window while maintaining performance.

Going Beyond the Context Window: Recursive Language Models in Action