arXiv cs.AI· Luke J. O'Connor·· 2 天前AI 评分40
语言模型智能体的自编程执行（SPE）与 Spell 语言
Self-Programmed Execution for Language-Model Agents
AI 导读
论文提出自编程执行（SPE），让模型补全本身充当编排程序，harness 只负责求值而不施加固定编排策略。为此作者设计了可自我编辑并重新求值的 Lisp 方言 Spell，使模型调用等副作用表达式在重求值时不重放。未经 SPE 或 Spell 训练的前沿模型可在上下文内学会 Spell 并完成挑战性智能体任务，代码已开源，论文被 NeurIPS 2026 接收。
正文 · 原文
Abstract:At the heart of a language model agent is an orchestrator program which is responsible for the state transition between consecutive turns. In existing systems, this program is implemented in the harness. This paper introduces self-programmed execution (SPE), in which the model completion is itself the orchestrator program, and the harness evaluates this program but does not impose any fixed orchestration policy. Formally, an SPE state of an agentic machine is one from which a model completion can load any successor state up to an embedding, and an instance of an SPE state is one which executes the model completion as a program. In practice, the program requires a language with unusual features, as its source code doubles as the content of the model's context window. I therefore introduce Spell, a Lisp dialect in which programs can edit and re-evaluate themselves, and effectful expressions like model invocations are structured such that re-evaluating an edited program does not replay its side effects. Experiments with existing models, not trained for SPE or Spell, show that frontier models can learn Spell in-context and accomplish challenging agentic tasks. These results demonstrate how a LM can act as an agent without any fixed orchestration policy, and they raise the question of what self-orchestration strategies might be learned by a model trained for self-programmed execution. Code is available at this https URL.
| Comments: | Camera-ready version; accepted at NeurIPS 2026. 77 pages, including appendices and checklist |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2605.06898 [cs.AI] |
| (or arXiv:2605.06898v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2605.06898 arXiv-issued DOI via DataCite |
Submission history
From: Luke O'Connor [view email]
[v1]
Thu, 7 May 2026 19:58:28 UTC (290 KB)
[v2]
Thu, 15 Oct 2026 15:37:58 UTC (326 KB)
来源：arXiv cs.AI · arxiv.org