The ๐ \mathbf{Y} -Combinator for LLMs: Solving Long-Context Rot with ฮป \lambda -Calculus
Amartya Roy, Rasul Tutunov, Xiaotong Ji, Matthieu Zimmer, Haitham Bou-Ammar ยท 2026
Read The ๐ \mathbf{Y} -Combinator for LLMs: Solving Long-Context Rot with ฮป \lambda -Calculus by Amartya Roy, Rasul Tutunov, Xiaotong Ji, Matthieu Zimmer, Haitham Bou-Ammar (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-20105.txt.
Description
Roy and co-authors give recursive language models a typed runtime based on lambda calculus. Long inputs strain the fixed context window of language models. Recursive models split the prompt and solve subproblems in pieces. Existing versions let the model write free control code in a read-eval-print loop. That freedom makes runs hard to verify or predict. The ฮป-RLM frame runs a small library of pre-verified combinators instead. Neural inference happens only on bounded leaf subproblems. It beat standard recursive models in 29 of 36 test comparisons.
Catalog record
| CREATOR | Amartya Roy, Rasul Tutunov, Xiaotong Ji, Matthieu Zimmer, Haitham Bou-Ammar |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 82000 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.20105