Research map / Time series / Foundation models for time series
Giving Sensors a Voice: Multimodal JEPA for Semantic Time-Series Embeddings
Utsav Dutta, Gerardo Pastrana, Sina Khoshfetrat Pakazad, Henrik Ohlsson
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace patch embedding with description-free TCN featurization (ablation) | CHARM · UEA classification (ablation) | accuracy: 63.0% -> 68.0% | Improved |
| add text-conditioned convolution kernels to the TCN (ablation) | CHARM · UEA classification (ablation) | accuracy: 68.0% -> 71.4% | Improved |
| add text-based kernel gating only to the TCN (ablation) | CHARM · UEA classification (ablation) | accuracy: 68.0% -> 69.0% | Improved |
| add text-conditioned convolutions to the TCN (ablation) | CHARM · ETTh1 forecasting, horizon 168 | MSE: 0.54 -> 0.42 | Improved |
| add text-conditioned convolutions to the TCN (ablation) | CHARM · ETTh2 forecasting, horizon 168 | MSE: 0.62 -> 0.57 | Improved |
| add channel gating G to attention (ablation) | CHARM · UEA classification (ablation) | accuracy: 67.9% -> 69.8% | Improved |
| add time-offset attention Delta (ablation) | CHARM · UEA classification (ablation) | accuracy: 67.9% -> 68.6% | Improved |
| combine Delta and G text attention (ablation) | CHARM · UEA classification (ablation) | accuracy: 67.9% -> 71.4% | Improved |
| add Delta only text attention (ablation) | CHARM · ETTh1 forecasting, horizon 168 | MSE: 0.46 -> 0.50 | Hurt |
| replace reconstruction loss with JEPA latent prediction (ablation) | CHARM · UEA classification | accuracy: +4.4pp | Improved |
| fine-tune encoder end-to-end instead of frozen embedding with SVM | CHARM · UEA multivariate classification | average accuracy: 79.6 -> 80.9 | Improved |
| end-to-end fine-tune with non-linear head instead of linear probe | CHARM · Weather forecasting (mean over horizons) | MSE: 0.230 -> 0.222 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 4 on most shared results.
- MOMENT — MOMENT: A Family of Open Time-series Foundation Models (ICML 2024) · better on 2 of 2 shared results
- VisionTS — VisionTS: Visual Masked Autoencoders Are Free-Lunch Zero-Shot Time Series Forecasters (ICML 2025) · better on 0 of 1 shared result
- TimesNet — TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis (ICLR 2023) · better on 1 of 1 shared result
- TS2Vec — TS2Vec: Towards Universal Representation of Time Series (AAAI 2022) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- TimeCMA: Towards LLM-Empowered Multivariate Time Series Forecasting via Cross-Modality Alignment (AAAI 2025 · similarity 0.85)
- Multi-Patch Prediction: Adapting Language Models for Time Series Representation Learning (ICML 2024 · similarity 0.84)
- $S^2$IP-LLM: Semantic Space Informed Prompt Learning with LLM for Time Series Forecasting (ICML 2024 · similarity 0.84)
- Frequency-Masked Embedding Inference: A Non-Contrastive Approach for Time Series Representation Learning (AAAI 2025 · similarity 0.84)
- VEDIT: Latent Prediction Architecture For Procedural Video Representation Learning (ICLR 2025 · similarity 0.84)
- TEMPO: Prompt-based Generative Pre-trained Transformer for Time Series Forecasting (ICLR 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
