mlsanityData to 7 Oct 2026

Which Histories Matter for Time Series Forecasting? Learning Predictive Relevance with Future Supervision

arXiv 2608.23221 ↗First posted 24 Aug 2026cs.IR cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Retrieval-augmented time-series forecasting typically selects historical examples by similarity between observed pasts, although similar pasts can evolve differently. We propose Predictive Relevance Retrieval (PRR), which uses realized future compatibility as privileged supervision to learn a retrieval function that remains strictly past-only at inference. PRR combines Pearson retrieval with a futuresupervised predictive representation to expand candidate support, then reranks the union using statistical and learned pair relations. Across six datasets and four long horizons, PRR improves Pearson retrieval in 23 of 24 conditions, reducing AnalogFutureMSE by 26.7% on average. A candidate-budget-matched variant, PRR-B100, retains nearly the same retrieval improvement while using at most 100 candidates at inference, showing that the gain is not explained simply by a larger candidate pool. We then connect the retriever to five frozen forecasting backbones using validation-calibrated trust. Downstream effects are heterogeneous, and calibration primarily reduces harmful retrieval use rather than making improved retrieval universally beneficial. These results show that predictive relevance and forecast utility are empirically distinct objectives.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 19 of its 25 references that are in the map. First posted on arXiv 24 Aug 2026.

Builds on

19 of 25 references.

References · 19

  1. 2022A Time Series is Worth 64 Words: Long-term Forecasting with Transformerscited by 1,981
  2. 2022Are Transformers Effective for Time Series Forecasting?cited by 1,400
  3. 2023iTransformer: Inverted Transformers Are Effective for Time Series Forecastingcited by 906
  4. 2024TimeMixer: Decomposable Multiscale Mixing for Time Series Forecastingcited by 333
  5. 2024Retrieval Augmented Time Series Forecastingcited by 40
  6. 2025TS-RAG: Retrieval-Augmented Generation based Time Series Foundation Models are Stronger Zero-Shot Forecastercited by 36
  7. 2024Learning Pattern-Specific Experts for Time Series Forecasting Under Patch-level Distribution Shiftcited by 23
  8. 2025Efficient Multivariate Time Series Forecasting via Calibrated Language Models with Privileged Knowledge Distillationcited by 20
  9. 2024Proactive Model Adaptation Against Concept Drift for Online Time Series Forecastingcited by 18
  10. 2025Improving Time Series Forecasting via Instance-aware Post-hoc Revisioncited by 14
  11. 2021Using Time-Series Privileged Information for Provably Efficient Learning of Prediction Modelscited by 6
  12. 2025Predicting the Future by Retrieving the Pastcited by 6
7 more
  1. 2025Nearest Neighbor Multivariate Time Series Forecastingcited by 3
  2. 2026Seg-MoE: Multi-Resolution Segment-wise Mixture-of-Experts for Time Series Forecasting Transformerscited by 3
  3. 2026Channel-wise Retrieval for Multivariate Time Series Forecastingcited by 3
  4. 2026Stationarity-Aware Retrieval-Augmented Time Series Forecastingcited by 2
  5. 2026Semantics-Enhanced Retrieval-Augmented Time Series Forecastingcited by 2
  6. 2026Learning to Query History: Nonstationary Classification via Learned Retrievalcited by 1
  7. 2026KReF: Training-Free Retrieval for Long-Term Time-Series Forecasting and Predictive Uncertaintycited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Retrieval Is Not Enough: Refreshing Memory for Frozen Time-Series Forecasters
  2. 2025Retrieval Augmented Time Series Forecasting
  3. 2026Factorize to Generalize: Retrieval-Guided Invariant-Dynamic Decomposition for Time Series Forecasting
  4. 2026Spectral Retrieval-Augmented Time-Series Forecasting
  5. 2025QuiZSF: A Retrieval-Augmented Framework for Zero-Shot Time Series Forecasting
  6. 2026SEER: Self-Evolving Event Reasoning and Retrieval for Time Series Forecasting
  7. 2026RATL: Learning from Retrieved Residuals for Robust Multivariate Time-Series Forecasting
  8. 2026Not All Retrievals are Useful: Cross-Attention for Input-Aware RAG in Time Series Forecasting