{
  "abstract": "Time Series Foundation Models (TSFMs) currently provide state-of-the-art results in forecasting tasks. They are available out-of-the-box and rely on in-context learning to make their predictions, which makes the quality of their performance highly sensitive to the user-selected lookback, covariates, horizon and training data distributions. In practise, the quality of the forecasts are variable but complementary, which highlights the need for a principled ensembling approach, rather than selecting the best context. This paper introduces Latent Inference-Time Guidance for TSFMs, which adaptively combines a pool of TSFM forecasts through a time-dependent latent space with independent components. The framework comes equipped with identifiability and reconstruction guarantees, whilst maintaining the off-the-shelf aspect of foundation models. We provide experiments on datasets at various frequencies and from multiple domains: these show that the approach is competitive with traditional ensembling approaches.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": "2609.38058",
  "author_total": 6,
  "authors": [
    {
      "name": "Chloé Hashimoto-Cullen"
    },
    {
      "name": "Amaury Durand"
    },
    {
      "name": "Laurent Bozzi"
    },
    {
      "name": "Benjamin Guedj"
    },
    {
      "name": "Yannig Goude"
    },
    {
      "name": "Sylvain Le Corff"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/a02vst9q8j2jq.bib",
  "cited_by": 0,
  "doi": "10.48550/arxiv.2609.38058",
  "fulltext_url": null,
  "id": "a02vst9q8j2jq",
  "issns": [],
  "landing_url": "https://hal.science/hal-05773148v1",
  "language": "en",
  "licence": "CC-BY-NC-SA-4.0",
  "licence_label": "CC BY-NC-SA 4.0",
  "licence_source": "article",
  "licence_url": "https://creativecommons.org/licenses/by-nc-sa/4.0/",
  "metadata_licence": "CC0-1.0",
  "native_subjects": [
    [
      "arxiv",
      "cs.LG"
    ],
    [
      "arxiv",
      "stat.ME"
    ],
    [
      "arxiv",
      "stat.ML"
    ],
    [
      "hal",
      "0.info"
    ],
    [
      "hal",
      "0.stat"
    ],
    [
      "openalex",
      "1711|Time Series Analysis and Forecasting"
    ]
  ],
  "openalex": "W7214943972",
  "pdf_url": "https://hal.science/hal-05773148/document",
  "pmcid": null,
  "pmid": null,
  "published": "2026-09-29",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/a02vst9q8j2jq.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:17:45Z",
      "native_id": "hal-05773148",
      "source": "hal",
      "url": "https://hal.science/hal-05773148v1"
    },
    {
      "first_seen": "2026-10-30T12:25:41Z",
      "native_id": "2609.38058",
      "source": "arxiv",
      "url": "https://arxiv.org/abs/2609.38058"
    }
  ],
  "subjects": [
    {
      "field": "mathematics",
      "origin": "arxiv:stat.ME",
      "subfield": "statistics"
    }
  ],
  "title": "Latent Inference-Time Guidance of Time Series Foundation Models",
  "title_text": "Latent Inference-Time Guidance of Time Series Foundation Models",
  "type": "preprint",
  "updated_at": "2026-10-30T12:25:41Z",
  "url": "https://research.draconitic.com/paper/a02vst9q8j2jq",
  "venue": "HAL",
  "versions": []
}