[KongchangAI]
ConceptRecursive Language Model / RLM

Recursive Language Model

A language model paradigm that decomposes complex problems into hierarchical sub-problems, each solved independently within its own context window. By isolating contexts, the approach prevents shortcut learning and is intended to achieve stronger out-of-distribution generalization and reasoning compared to standard single-context models.

Timeline (last 90 days)

Aug 25

Prime Agent不提供固定工具集,而是提供一个持久化的iPython内核,工具、技能和子智能体都是模型可直接编写运行的Python代码,这一设计称为递归语言模型(RLM)

Unverified50%

All Facts (1)

Source Articles