On the Divergence of Accuracy and Mechanism Consistency in Time Series World Models
作者: Haochen Zhang, Jiaheng Guo, Zhen Xu, Zachary Plotkin, Nicholas Konz, Zhen Tan, Tianlong Chen
分类: cs.AI
发布日期: 2026-10-01
💡 一句话要点
提出时间序列世界模型以解决预测准确性与机制一致性问题
🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture)
关键词: 时间序列预测 机制一致性 模型评估 智能控制 医疗监测 工程基础设施 深度学习
📋 核心要点
- 现有时间序列预测方法在未执行计划的情况下,无法有效评估其对变化计划的响应,导致准确性与机制一致性之间的分歧。
- 本文提出了一种新的形式化方法,区分状态、行动和外部输入,并引入机制一致性度量,以评估预测模型的设计选择。
- 实验结果表明,冻结潜在空间和输出侧融合在所有数据集上平均降低MAE,方向性监督显著提高机制一致性,且MAE未受影响。
📝 摘要(中文)
时间序列世界模型(TSWM)旨在从观察历史、计划行动和外部输入中预测受控系统的状态。现有方法在执行计划下训练和评估预测误差,但未测试未执行计划的响应。本文通过形式化和基准测试探讨设计选择的重要性及准确预测者对变化计划的响应。形式化将状态、行动和外部输入分离,提出机制一致性这一度量,检查行动变化是否按预期方向影响预测。基准测试整合了八个公共数据集,结果显示冻结潜在预测空间和输出融合显著降低平均绝对误差(MAE),而方向性监督则提高机制一致性。整体为TSWM提供了准确性和一致性的训练方案。
🔬 方法详解
问题定义:本文旨在解决时间序列预测模型在未执行计划情况下的响应评估问题。现有方法未能有效比较不同计划的预测结果,导致准确性与机制一致性之间的分歧。
核心思路:论文提出了一种形式化的框架,明确区分状态、行动和外部输入,并引入机制一致性度量,旨在确保预测模型能够准确反映行动变化对系统状态的影响。
技术框架:整体架构包括数据预处理、模型训练和评估三个主要模块。数据预处理阶段整合了多个公共数据集,模型训练阶段采用不同的网络结构和损失函数,评估阶段则关注预测误差和机制一致性。
关键创新:最重要的创新在于引入机制一致性度量,能够量化预测模型对行动变化的响应是否符合预期方向。这一设计与传统方法的本质区别在于关注模型的解释性和可用性。
关键设计:在模型训练中,采用冻结潜在空间和输出侧融合的策略,显著降低MAE。同时,方向性监督作为损失函数的一部分,专门惩罚预测响应的错误方向,从而提高机制一致性。具体参数设置和网络结构在实验中进行了详细比较。
🖼️ 关键图片
📊 实验亮点
实验结果显示,冻结潜在预测空间平均降低MAE 9.9%,输出侧融合降低12.7%。此外,方向性监督显著提高机制一致性,而MAE保持不变。这些结果表明,提出的方法在多个数据集上均表现出色,具有较强的实用价值。
🎯 应用场景
该研究的潜在应用领域包括智能控制系统、医疗监测和工程基础设施管理等。通过提高时间序列预测模型的准确性和一致性,该方法能够为实时决策提供更可靠的支持,进而提升系统的整体性能和安全性。
📄 摘要(原文)
A time series world model (TSWM) predicts a controlled system's state from its observed history and planned actions and exogenous inputs. Current approaches build forecasters with actions as covariates, trained and evaluated on prediction error under the executed plan. Yet world models compare unexecuted plans, but their responses to changed plans remain untested. We ask which design choices matter and whether accurate forecasters respond to changed plans as real systems do. We address both with a formalization and benchmark. The formalization separates state, actions and exogenous inputs, distinguishes continuous, mode and event actions, and introduces mechanism consistency, a metric built on declared action-state relations with known directions, such as a vasopressor raising blood pressure: it checks whether shifting an action moves the forecast in the declared direction. The benchmark consolidates eight public datasets with real actions from engineered infrastructure and clinical care, varying prediction space, plan fusion and plan encoding across seven backbones and five seeds. First, a frozen latent prediction space lowers MAE by 9.9% over observation space and gated output fusion lowers it by 12.7% over input concatenation on average, with both improving all eight datasets; temporal plan encoding changes average MAE by at most 2.2%. Second, prediction error and mechanism consistency diverge: the lowest-error configuration is at or below chance in consistency on four of five datasets with declared mechanisms, and no design choice avoids this. Finally, directional supervision, a loss penalizing the wrong-signed part of the response to a shifted action, significantly raises consistency on penalized mechanisms with no change in MAE. Together they give TSWMs a recipe: a frozen latent space and output-side fusion for accuracy, and a training objective for mechanism consistency.