Update README paper.

Updated reference to the NeurIPS 2026 paper in the README.

Author: alexzhang13

README.md

@@ -32,7 +32,7 @@ Recursive Language Models (RLMs) are a task-agnostic inference paradigm for lang
 
 RLMs are a bet on future "language model" design choices. We argue for a [CodeAct](https://arxiv.org/abs/2402.01030)-style harness (i.e. all language models should have access to a code environment) with sub-(R)LM calls as functions in code, and context / prompts as objects in code. RLMs explicitly defer code execution with sub-calls as functions to the language model itself, which is incredibly flexible and lends itself well to scale if trained correctly. We want to move away from the JSON tool-calling standard for both sub-agents and generic tool calls. The naming comes from the fact that such a system is itself a "language model" (a probabilistic mapping from text to text) that builds around and relies on recursive sub-LLM calls.
 
-This repository provides both an extensible inference engine and training environment for using RLMs around standard API-based and local LLMs. The initial experiments and idea were proposed in a [blogpost](https://alexzhang13.github.io/blog/2025/rlm/) in 2025, with expanded results in an [arXiv preprint](https://arxiv.org/abs/2512.24601).
+This repository provides both an extensible inference engine and training environment for using RLMs around standard API-based and local LLMs. The initial experiments and idea were proposed in a [blogpost](https://alexzhang13.github.io/blog/2025/rlm/) in 2025, with expanded results in our [NeurIPS 2026 paper](https://arxiv.org/abs/2512.24601).
 
 We now also include a [verifiers](https://github.com/PrimeIntellect-ai/verifiers) training environment based on Prime Intellect's [prime-rl](https://github.com/PrimeIntellect-ai/prime-rl) in the `training/` folder. Train your own RLMs, which directly can be plugged into our inference engine!