Replaces unstable free-form recursive LLM code with a typed functional runtime grounded in lambda-calculus.
March 23, 2026
Original Paper
The $\mathbf{Y}$-Combinator for LLMs: Solving Long-Context Rot with $λ$-Calculus
arXiv · 2603.20105
The Takeaway
Addresses the 'long-context rot' problem where recursive reasoning fails over time. By using pre-verified combinators and symbolic control flow, it achieves massive accuracy gains (+21 points) and significant latency reductions.
From the abstract
LLMs are increasingly used as general-purpose reasoners, but long inputs remain bottlenecked by a fixed context window. Recursive Language Models (RLMs) address this by externalising the prompt and recursively solving subproblems. Yet existing RLMs depend on an open-ended read-eval-print loop (REPL) in which the model generates arbitrary control code, making execution difficult to verify, predict, and analyse.We introduce $\lambda$-RLM, a framework for long-context reasoning that replaces free-f