如预期？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
Qingfu Zhu · Huanyang Zheng · Ting Liu · Wanxiang Che
中文摘要
LLM 规划常依赖反应式执行器（如 ReAct），每步只生成单个动作以保持对最新观察的接地，推理成本高。批量执行器在单次模型调用中预测多个动作以降低成本，但因忽略中间观察而易产生累积误差。提出面向批量执行器的期望-观察一致性验证（EOCV）机制，在保持接地的同时提升效率。该机制生成动作结果的期望并与实际观察比较：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。在多个骨干模型上的文本与多模态基准实验显示，EOCV 与反应式执行器的平均任务性能大体相当，平均加权 token 用量减少 29-35%。该机制与模型无关，集成到先进规划器后仍有效。
关键要点
- 01问题：批量执行器忽略中间观察，易产生累积误差，而反应式执行器推理成本高。
- 02方法：生成动作结果期望并与观察比对，一致则继续，不匹配则终止并重新规划。
- 03结果：在文本与多模态基准上平均任务性能与反应式执行器大体相当。
- 04结果：平均加权 token 用量减少 29-35%，且机制与模型无关。
解读
尚无解读。
原始英文摘要
arXiv:2611.02438v1 Announce Type: new 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.