Self-Programmed Execution for Language-Model Agents

TL;DR

This paper proposes self-programmed execution (SPE), enabling language models to act as agents without fixed turn-based policies, using a novel Lisp-like language called Spell for self-editing and re-evaluation.

Contribution

It introduces SPE and Spell, allowing language models to self-orchestrate their actions, and demonstrates their effectiveness on challenging agentic tasks.

Findings

Frontier models can operate under SPE without training for it.

Models can perform complex agentic tasks using self-programmed execution.

Code implementation is available at the provided GitHub link.

Abstract

At the heart of existing language model agents is a fixed orchestrator program responsible for the state transition between consecutive turns. This paper introduces self-programmed execution (SPE), an agent architecture in which the model completion is itself the orchestrator program, and the harness evaluates this program but does not impose its own orchestration policy. I formalize this idea using agentic machines: an SPE state is one from which a model completion can load any state of an embedded copy of the machine, meaning that it is subject to no fixed turn-to-turn orchestration policy. Realizing SPE in practice is nontrivial because the same data is both model context and executable program. I therefore introduce Spell, a Lisp-based language 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 operate in this regime and accomplish challenging agentic tasks. These results demonstrate how an 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 https://github.com/lukejoconnor/spell .