Pith. sign in

REVIEW 9 cited by

TimeCMA: Towards LLM-Empowered Multivariate Time Series Forecasting via Cross-Modality Alignment

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2406.01638 v5 pith:KKELCZJZ submitted 2024-06-03 cs.LG cs.AIcs.CL

Chenxi Liu , Qianxiong Xu , Hao Miao , Sun Yang , Lingzheng Zhang , Cheng Long , Ziyue Li , Rui Zhao This is my paper · ORCID
classification cs.LGcs.AIcs.CL
keywords seriestimeembeddingspromptalignmentcross-modalitydisentangledencoding
verification ladder T0 review T1 audit T2 compute T3 formal

Signed reviews

No signed human review yet.

0 comments
read the original abstract

Multivariate time series forecasting (MTSF) aims to learn temporal dynamics among variables to forecast future time series. Existing statistical and deep learning-based methods suffer from limited learnable parameters and small-scale training data. Recently, large language models (LLMs) combining time series with textual prompts have achieved promising performance in MTSF. However, we discovered that current LLM-based solutions fall short in learning disentangled embeddings. We introduce TimeCMA, an intuitive yet effective framework for MTSF via cross-modality alignment. Specifically, we present a dual-modality encoding with two branches: the time series encoding branch extracts disentangled yet weak time series embeddings, and the LLM-empowered encoding branch wraps the same time series with text as prompts to obtain entangled yet robust prompt embeddings. As a result, such a cross-modality alignment retrieves both disentangled and robust time series embeddings, "the best of two worlds", from the prompt embeddings based on time series and prompt modality similarities. As another key design, to reduce the computational costs from time series with their length textual prompts, we design an effective prompt to encourage the most essential temporal information to be encapsulated in the last token: only the last token is passed to downstream prediction. We further store the last token embeddings to accelerate inference speed. Extensive experiments on eight real datasets demonstrate that TimeCMA outperforms state-of-the-arts.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 9 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Efficient Test-Time Scaling for LLM-based Time Series Forecasting

    cs.LG 2026-08 conditional novelty 6.0 of 10

    SCALER predicts a coarse future shape with a lightweight Transformer, then lets a frozen LLM refine it in fixed steps, improving forecast accuracy and cutting inference cost versus test-time scaling baselines.

  2. NSW-EPNews: A News-Augmented Benchmark for Electricity Price Forecasting with LLMs

    cs.LG 2025-05 reject novelty 6.0 of 10

    LLMs forecast electricity prices worse than ARIMA on the new NSW-EPNews benchmark and frequently hallucinate by echoing, offsetting, or repeating historical prices.

  3. Context-Aware Probabilistic Modeling with LLM for Multimodal Time Series Forecasting

    cs.LG 2025-05 conditional novelty 6.0 of 10

    CAPTime aligns time series patches with summarized text via cross-attention and decodes forecasts from a mixture of distribution experts through a frozen GPT-2, reporting improved multimodal forecasting accuracy.

  4. Retrieval-augmented Large Language Models for Financial Time Series Forecasting

    cs.CL 2025-02 reject novelty 6.0 of 10

    A financial time-series retriever trained on StockLLM's own confidence scores improves that same StockLLM's next-day up/down prediction accuracy on three datasets by about 1 to 3 percentage points.

  5. BALM-TSF: Balanced Multimodal Alignment for LLM-Based Time Series Forecasting

    cs.AI 2025-08 conditional novelty 5.0 of 10

    BALM-TSF combines a statistical-prompt text branch with a patch-based time series branch, using scaling plus contrastive alignment to balance the two modalities, improving long-term and few-shot forecasting on five of...

  6. Can LLM Improve for Expert Forecast Combination? Evidence from the European Central Bank Survey

    stat.AP 2025-06 reject novelty 5.0 of 10

    A zero-shot LLM prompt beats equal-weighted averaging for one-year ECB SPF forecasts in one regression, but the result is fragile, the comparison is asymmetric, and no code or data are provided.

  7. Dual-Forecaster: A Multimodal Time Series Model Integrating Descriptive and Predictive Texts

    cs.LG 2025-05 conditional novelty 5.0 of 10

    Dual-Forecaster demonstrates that jointly using historical descriptive text and future predictive text improves time series forecasting accuracy over text-free and single-text baselines.

  8. MoFE-Time: Mixture of Frequency Domain Experts for Time-Series Forecasting Models

    cs.LG 2025-07 conditional novelty 4.0 of 10

    MoFE-Time reports average MSE 0.2755 and MAE 0.3226 across six public benchmarks, about 7% lower than Time-MoE, by adding frequency-domain experts to a Mixture of Experts transformer.

  9. Adversarial Vulnerabilities in Large Language Models for Time Series Forecasting

    cs.LG 2024-12 conditional novelty 4.0 of 10

    A simple gradient-free 'directional gradient approximation' attack makes LLM time series forecasters degrade more than equivalent random noise, across GPT-3.5, GPT-4, LLaMa, Mistral, TimeGPT, and TimeLLM.

Pith tools