mlsanityData to 14 Oct 2026

Tracing the Evidence Behind Zero-Shot Time-Series Forecasting: A Source-First Taxonomy and Audit Framework

arXiv 2609.21425 ↗First posted 18 Sep 2026cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Zero-shot time-series forecasting (TSF) is often described as forecasting without target-specific parameter updates, but that training-status condition does not specify what evidence the system may use. A frozen language model prompted with serialized values, a time-series model pretrained on broad forecasting corpora, and a retrieval-augmented forecaster may all satisfy the no-update condition while drawing on different transferable evidence. This paper argues that zero-shot TSF should therefore be governed as an evidence-access claim. We propose a source-first taxonomy that separates three primary evidence sources---frozen LLM prior reuse, parametric time-series pretraining, and retrieval-augmented external memory---from the architectures that implement them. After the source is identified, four additional audit questions remain: task interface, forecast object and scoring, prediction-time context, and resource budget. The resulting agenda is to make zero-shot leaderboards auditable by reporting evidence boundaries and interface assumptions alongside scores, so that benchmark progress reflects transferable forecasting capability rather than undisclosed changes in context, memory, or budget.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 12 of its 22 references that are in the map. First posted on arXiv 18 Sep 2026.

Builds on

12 of 22 references.

References · 12

  1. 2024Unified Training of Universal Time Series Forecasting Transformerscited by 614
  2. 2023Large Language Models Are Zero-Shot Time Series Forecasterscited by 500
  3. 2024Time-MoE: Billion-Scale Time Series Foundation Models with Mixture of Expertscited by 287
  4. 2024GIFT-Eval: A Benchmark For General Time Series Forecasting Model Evaluationcited by 212
  5. 2019Show Your Work: Improved Reporting of Experimental Resultscited by 169
  6. 2025TiRex: Zero-Shot Forecasting Across Long and Short Horizons with Enhanced In-Context Learningcited by 120
  7. 2023Large Models for Time Series and Spatio-Temporal Data: A Survey and Outlookcited by 110
  8. 2024ChatTime: A Unified Multimodal Time Series Foundation Model Bridging Numerical and Textual Datacited by 100
  9. 2024LSTPrompt: Large Language Models as Zero-Shot Time Series Forecasters by Long-Short-Term Promptingcited by 59
  10. 2025TS-RAG: Retrieval-Augmented Generation based Time Series Foundation Models are Stronger Zero-Shot Forecastercited by 40
  11. 2025Rethinking Evaluation in the Era of Time Series Foundation Models: (Un)known Information Leakage Challengescited by 16
  12. 2025How Foundational are Foundation Models for Time Series Forecasting?cited by 5

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Evaluating System 1 vs. 2 Reasoning Approaches for Zero-Shot Time Series Forecasting: A Benchmark and Insights
  2. 2026Foundations without Fundamentals: Zero-Shot Blind Spots in Time Series FMs
  3. 2025QuiZSF: A Retrieval-Augmented Framework for Zero-Shot Time Series Forecasting
  4. 2026Not All Retrievals are Useful: Cross-Attention for Input-Aware RAG in Time Series Forecasting
  5. 2024TimeRAF: Retrieval-Augmented Foundation model for Zero-shot Time Series Forecasting
  6. 2026Zero-shot Forecasting by Simulation Alone
  7. 2026GRAM: Correcting Frozen Time-Series Foundation Models via Graph-Retrieved Amplitude Memory
  8. 2026Retrieval Is Not Enough: Refreshing Memory for Frozen Time-Series Forecasters