mlsanityData to 14 Oct 2026

Overcoming the Modality Gap in Context-Aided Forecasting

arXiv 2603.12451 ↗First posted 12 Mar 2026cs.LGcited by 5

Not on the map. Why?

Abstract (arXiv)

Context-aided forecasting (CAF) holds promise for integrating domain knowledge and forward-looking information, enabling AI systems to surpass traditional statistical methods. However, recent empirical studies reveal a puzzling gap: multimodal models often fail to outperform their unimodal counterparts. We hypothesize that this underperformance stems from poor context quality in existing datasets, as verification is challenging. To address these limitations, we introduce a semi-synthetic data augmentation method that generates contexts both descriptive of temporal dynamics and verifiably complementary to numerical histories. This approach enables massive-scale dataset creation, resulting in CAF-7M, a corpus of 7 million context-augmented time series windows, including a rigorously verified test set. We demonstrate that semi-synthetic pre-training transfers effectively to real-world evaluation, and show clear evidence of context utilization. Our results suggest that dataset quality, rather than architectural limitations, has been the primary bottleneck in context-aided forecasting.

Key facts. Cited by 5 papers in the map (data through 14 Oct 2026). Builds on 27 of its 38 references that are in the map. First posted on arXiv 12 Mar 2026.

Builds on

27 of 38 references.

References · 27

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2025Qwen3 Technical Reportcited by 8,820
  3. 2024Chronos: Learning the Language of Time Seriescited by 759
  4. 2023Time-LLM: Time Series Forecasting by Reprogramming Large Language Modelscited by 700
  5. 2023Large Language Models Are Zero-Shot Time Series Forecasterscited by 500
  6. 2025Layer by Layer: Uncovering Hidden Representations in Language Modelscited by 311
  7. 2024GIFT-Eval: A Benchmark For General Time Series Forecasting Model Evaluationcited by 212
  8. 2025Chronos-2: From Univariate to Universal Forecastingcited by 194
  9. 2022PromptCast: A New Prompt-based Learning Paradigm for Time Series Forecastingcited by 172
  10. 2024Are Language Models Actually Useful for Time Series Forecasting?cited by 168
  11. 2023UniTime: A Language-Empowered Unified Model for Cross-Domain Time Series Forecastingcited by 144
  12. 2024TimeMixer++: A General Time Series Pattern Machine for Universal Predictive Analysiscited by 116
15 more
  1. 2024ChatTime: A Unified Multimodal Time Series Foundation Model Bridging Numerical and Textual Datacited by 100
  2. 2024From News to Forecast: Integrating Event Analysis in LLM-Based Time Series Forecasting with Reflectioncited by 98
  3. 2024Language Models Still Struggle to Zero-shot Reason about Time Seriescited by 88
  4. 2024Context is Key: A Benchmark for Forecasting with Essential Textual Informationcited by 87
  5. 2024Time-MMD: Multi-Domain Multimodal Dataset for Time Series Analysiscited by 74
  6. 2024AutoTimes: Autoregressive Time Series Forecasters via Large Language Modelscited by 65
  7. 2025Time-VLM: Exploring Multimodal Vision-Language Models for Augmented Time Series Forecastingcited by 63
  8. 2024LLM Processes: Numerical Predictive Distributions Conditioned on Natural Languagecited by 55
  9. 2025MTBench: A Multimodal Time Series Benchmark for Temporal Reasoning and Question Answeringcited by 55
  10. 2025Aurora: Towards Universal Generative Multimodal Time Series Forecastingcited by 29
  11. 2024Multi-Patch Prediction: Adapting LLMs for Time Series Representation Learningcited by 25
  12. 2025T2S: High-resolution Time Series Generation with Text-to-Series Diffusion Modelscited by 20
  13. 2025FinMultiTime: A Four-Modal Bilingual Dataset for Financial Time-Series Analysiscited by 11
  14. 2025When Does Multimodality Lead to Better Time Series Forecasting?cited by 11
  15. 2025Dual-Forecaster: A Multimodal Time Series Model Integrating Descriptive and Predictive Textscited by 3

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026TimeBraid: Unifying Time Series and Language for Understanding and Forecastingcited by 0
  2. 2026OpenTSLM TeeMoE: A Unified Time-Series Language Model for Forecasting, Contextual Prediction, and Reasoningcited by 0
  3. 2026TIMBRE: Teaching Time Series Forecasters to Read, Remember, and Reconcilecited by 0
  4. 2026Pythia: Toward Foundation World Models for Multimodal Time Seriescited by 0
  5. 2026Omni-Modal LLMs Know More Than They Saycited by 0