mlsanityData to 14 Oct 2026

RAID: Semantic Graph Diffusion for True Cold-Start and Cross-Lingual Forecasting

arXiv 2606.16925 ↗First posted 15 Jun 2026cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Time-series foundation models show strong transfer performance when given a non-empty history window. However, true cold-start scenarios, where a new item has no prior observations, violate this assumption. We propose RAID (Retrieval-Augmented Iterative Diffusion) a framework, which replaces history-based correlation learning with metadata-driven semantic retrieval and graph-conditioned diffusion. RAID maps textual metadata into a shared semantic space using a frozen multilingual embedding model and constructs an inductive retrieval graph that extends naturally to unseen items. It first forms a base forecast by aggregating information from semantically related neighbors, then refines this forecast with a gated diffusion module to model residual uncertainty. Under a strict true cold-start protocol, RAID outperforms strong foundation models and competitive baselines on both forecasting accuracy and prediction interval coverage, while reducing inference latency by an order of magnitude through non-autoregressive decoding. The shared semantic space also enables zero-shot cross-lingual transfer, allowing a model trained on English descriptions to generalize to items described in other languages without direct supervision.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 27 of its 34 references that are in the map. First posted on arXiv 15 Jun 2026.

Builds on

27 of 34 references.

References · 27

  1. 2020Denoising Diffusion Probabilistic Modelscited by 21,076
  2. 2023GPT-4 Technical Reportcited by 18,396
  3. 2024The Llama 3 Herd of Modelscited by 14,715
  4. 2017Graph Attention Networkscited by 8,923
  5. 2022Scalable Diffusion Models with Transformerscited by 6,054
  6. 2024Qwen2.5 Technical Reportcited by 4,942
  7. 2023Mistral 7Bcited by 3,164
  8. 2017FiLM: Visual Reasoning with a General Conditioning Layercited by 2,333
  9. 2024M3-Embedding: Multi-Linguality, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillationcited by 1,073
  10. 2017DeepAR: Probabilistic Forecasting with Autoregressive Recurrent Networkscited by 1,013
  11. 2021How Attentive are Graph Attention Networks?cited by 835
  12. 2019Temporal Fusion Transformers for Interpretable Multi-horizon Time Series Forecastingcited by 761
15 more
  1. 2024Chronos: Learning the Language of Time Seriescited by 759
  2. 2023A decoder-only foundation model for time-series forecastingcited by 692
  3. 2024Unified Training of Universal Time Series Forecasting Transformerscited by 614
  4. 2021CSDI: Conditional Score-based Diffusion Models for Probabilistic Time Series Imputationcited by 582
  5. 2020Connecting the Dots: Multivariate Time Series Forecasting with Graph Neural Networkscited by 566
  6. 2021Autoregressive Denoising Diffusion Models for Multivariate Probabilistic Time Series Forecastingcited by 337
  7. 2023Long-term Forecasting with TiDE: Time-series Dense Encodercited by 332
  8. 2024GIFT-Eval: A Benchmark For General Time Series Forecasting Model Evaluationcited by 212
  9. 2024Tiny Time Mixers (TTMs): Fast Pre-trained Models for Enhanced Zero/Few-Shot Forecasting of Multivariate Time Seriescited by 127
  10. 2023Lag-Llama: Towards Foundation Models for Probabilistic Time Series Forecastingcited by 113
  11. 2024Context is Key: A Benchmark for Forecasting with Essential Textual Informationcited by 87
  12. 2020Meta-learning framework with applications to zero-shot time-series forecastingcited by 70
  13. 2024In-Context Fine-Tuning for Time-Series Foundation Modelscited by 35
  14. 2025LangTime: A Language-Guided Unified Model for Time Series Forecasting with Proximal Policy Optimizationcited by 22
  15. 2024Graph Neural Patching for Cold-Start Recommendationscited by 3

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025TS-RAG: Retrieval-Augmented Generation based Time Series Foundation Models are Stronger Zero-Shot Forecaster
  2. 2024Retrieval-Augmented Generation Meets Data-Driven Tabula Rasa Approach for Temporal Knowledge Graph Forecasting
  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. 2026Beyond Similarity: Coverage-Aware Prompt Selection for Time Series Forecasting with LLMs
  6. 2025STELLA: Guiding Large Language Models for Time Series Forecasting with Semantic Abstractions
  7. 2026Semantics-Enhanced Retrieval-Augmented Time Series Forecasting
  8. 2023GenTKG: Generative Forecasting on Temporal Knowledge Graph with Large Language Models