arXiv cs.CL· Qingfu Zhu, Huanyang Zheng, Ting Liu, Wanxiang Che·· 11 小时前AI 评分34
通过期望-观察一致性验证实现高效 LLM 规划
As Expected? Efficient LLM Planning via Expectation-Observation Consistency Verification
AI 导读
研究者提出期望-观察一致性验证（EOCV）机制，让批量执行器在单次调用中预测多个动作时先生成动作结果期望，并与实际观察比对：一致则继续执行，不匹配则提前终止并从当前观察重新规划。在文本与多模态基准、多个基座模型上，EOCV 与 ReAct 等反应式执行器平均任务表现大体相当，平均加权 token 用量降低 29-35%。该机制与模型无关，接入现有先进规划器后仍然有效。
正文 · 原文
Abstract:LLM planning often relies on reactive executors (e.g., ReAct) that generate a single action at each step to remain grounded in the latest observations, incurring substantial inference cost. Batched executors reduce this cost by predicting multiple actions in a single model invocation, yet they are prone to compounding errors, as intermediate observations are ignored. We propose an expectation-observation consistency verification (EOCV) mechanism for batched executors, preserving grounding while improving efficiency. The mechanism generates expectations of action outcomes and compares them with the resulting observations. Consistency permits continued execution, whereas mismatches trigger early termination and replanning from the current observation. Experiments on text-based and multimodal benchmarks across multiple backbones show broadly comparable average task performance between EOCV and reactive executors, with a 29-35% reduction in average weighted token usage. Moreover, the mechanism is model-agnostic and remains effective when integrated into state-of-the-art planners.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.02438 [cs.CL] |
| (or arXiv:2611.02438v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.02438 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Huanyang Zheng [view email]
[v1]
Mon, 2 Nov 2026 02:59:57 UTC (2,009 KB)
来源：arXiv cs.CL · arxiv.org