LiveHouse-TS Introduces Open-World Living Benchmark for Time Series Foundation Models

Time Series Foundation Models (TSFMs) are gaining traction as a viable method for cross-domain zero-shot forecasting. However, current assessments depend on static benchmarks that utilize fixed historical test windows, which only provide average performance metrics and overlook the dynamics of real-world scenarios characterized by seasonal changes, distribution shifts, and unforeseen occurrences. To address this limitation, LiveHouse-TS has been introduced as the inaugural open-world living benchmark system for TSFMs. Unlike static benchmarks, it evaluates models prequentially using actual future data in open-world contexts, thereby transforming time series evaluation from mere snapshot accuracy to ongoing temporal relevance. This continuous infrastructure, rather than a one-time leaderboard, is detailed in an arXiv publication under ID 2608.17299.