arXiv：cs.AI（全量分类）· Haoyang Su, Weiran Huang·· 5 小时前AI 评分34
JevSpawn：通过组合式动作空间实现自适应智能体推理
JevSpawn: Adaptive Agentic Inference through Compositional Action Spaces
AI 导读
JevSpawn 是一种组合式策略，将自然语言任务描述与有限域概率探索连接起来，通过并行动作生成配合反馈驱动的分支选择、表示修正与备选方案恢复，在无需额外训练的情况下减少重复生成与上下文计算。在八项基准任务上对比七种智能体基线及一个 TypeSafe Jev 变体，JevSpawn 展现出更优的任务表现与更快的导航速度。
正文
Abstract:LLM agents generate intermediate reasoning and actions token by token, making extended interactions slow and computationally expensive. Jev-style models offer fast probabilistic predictions over finite fields, but require those fields to be specified in advance. This requirement limits autonomous task solving, where the available actions must be derived from natural language instructions and adapted through interaction. We introduce JevSpawn, a compositional policy that connects natural language task specifications to finite probabilistic exploration. Parallel action spawning is coupled with feedback driven branch selection, representation revision, and recovery from retained alternatives. Shared action structure and model prefixes reduce repeated generation and context computation without additional training. Evaluations on eight benchmark tasks against seven agent baselines and a TypeSafe Jev variant establish JevSpawn as a promising approach to structured agentic inference, with improved task performance and faster navigation.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.00437 [cs.AI] |
| (or arXiv:2610.00437v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.00437 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Haoyang Su [view email]
[v1]
Wed, 30 Sep 2026 17:23:01 UTC (4,164 KB)
来源：arXiv：cs.AI（全量分类） · arxiv.org