论文:Wasserstein-1 严重度评分无法预测时序基础模型的退化幅度
Severe but Harmless: Stress-Testing Shift Severity for Frozen Time-Series Forecasters
做时序预测或异常检测的朋友值得看:论文用3个交通数据集9种偏移证明Wasserstein-1这类严重度评分和模型实际退化根本对不上,最低分场景反而退化29%。如果你在用零样本TSFM选模型或做监控,这套等变性压力测试思路能直接借鉴。
一篇 arXiv 论文系统检验了分布偏移严重度指标对零样本时序基础模型(TSFM)的预测能力。实验在三个标准交通数据集上构造九种 regime shift,用三种主流严重度度量(含 Wasserstein-1)评分后发现:评分最高的偏移并非模型退化最严重的偏移,而得分最低的 regime 反而出现高达 29% 的性能退化。
论文指出,实例归一化(instance normalization)是造成这一脱钩的关键机制——作者用基于等变性(equivariance)的压力测试分析解释了为什么统计意义上的偏移严重度与模型实际退化之间在多数情况下不存在正相关。
对使用 TSFM 做异常检测或预测的团队而言,这意味着不能用 Wasserstein-1 等度量来预判哪些数据段会拖垮模型,需要转向直接的退化审计方法。
Severe but Harmless: Stress-Testing Shift Severity for Frozen Time-Series Forecasters
Zero-shot time-series foundation models (TSFMs) have become attractive alternatives to task-specific models because they handle diverse downstream tasks, including anomaly detection and forecasting, without per-series training. However, time-series data undergo distribution shifts, and the dissimilarity between two regimes is commonly scored with statistical measures such as Wasserstein-1, where a higher score is read as a more severe shift. In systematic experiments using three established severity measures on nine regime shifts from three standard traffic datasets, we show that these measures do not reliably predict degradation for instance-normalized TSFMs. The regimes rated most severe are not the regimes that degrade the models most, and degradation reaches 29% in the lowest-scoring regime. In most cases, the audit shows no positive relationship between measured severity and model degradation. We provide an equivariance-based stress-testing analysis that explains why severity readings can become decoupled from model degradation.
This report is categorized under AITOP's "Papers" desk. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include time series, Wasserstein-1, TSFM, distribution shift.
Source attribution: curated from arXiv cs.LG (@Korede R. Bishi, Dominic Okonkwo, Adetayo Okunoye).