Research map / Time series / Foundation models for time series
SciTS: Scientific Time Series Understanding and Generation with LLMs
Wen Wu, Ziyang Zhang, Liwei Liu, Xuenan Xu, Jimin Zhuang, Ke Fan, Qitan Lv, Junlin Liu, Chen Zhang, Zheqi Yuan, Siyuan Hou, Tianyi Lin and 3 more
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 |
|---|---|---|---|
| add the reprogramming module instead of a simple MLP (ablation) | TimeOmni · SciTS | performance | Improved |
| add patch expert routing instead of a fixed patch size (ablation) | TimeOmni · SciTS | performance | Improved |
| finetune Qwen2.5-VL-7B on TimeOmni training data | Qwen2.5-VL-7B · SciTS | task performance | No clear effect |
| finetune TimeMoE on TimeOmni training data | TimeMoE-Large · SciTS forecasting | performance | Mixed |
| feed time series as images instead of text | multimodal LLMs · SciTS understanding | average rank: Gemini2.5-Flash text 5.5 vs image 5.8 (GPT-5-mini 6.0 vs GPT-4.1-mini 6.1) | Improved |
| feed time series as text instead of images | LLMs · SciTS generation | average rank: Gemini2.5-Flash text 4.6 vs image 9.8 | Improved |
| use a dedicated time series encoder with LLM instead of general LLM only | TimeOmni vs Gemini2.5-Flash · SciTS understanding | average rank: 5.5 -> 1.9 | Improved |
| use a dedicated time series encoder with LLM instead of general LLM only | TimeOmni vs Gemini2.5-Flash · SciTS generation | average rank: 4.6 -> 4.1 | Improved |
| use specialised time series model (UniTS) instead of LLM | UniTS vs GPT-4.1-mini · SciTS understanding | average rank: 6.1 vs 7.9 | No clear effect |
| enable thinking mode in closed-source LLMs | closed-source LLMs · SciTS | performance | No clear effect |
Closest papers by meaning
- Can LLMs Understand Time Series Anomalies? (ICLR 2025 · similarity 0.90)
- STEM-LTS: Integrating Semantic-Temporal Dynamics in LLM-driven Time Series Analysis (AAAI 2025 · similarity 0.89)
- Time-LLM: Time Series Forecasting by Reprogramming Large Language Models (ICLR 2024 · similarity 0.89)
- Position: What Can Large Language Models Tell Us about Time Series Analysis (ICML 2024 · similarity 0.88)
- ChatTime: A Unified Multimodal Time Series Foundation Model Bridging Numerical and Textual Data (AAAI 2025 · similarity 0.88)
- TimeXL: Explainable Multi-modal Time Series Prediction with LLM-in-the-Loop (NeurIPS 2025 · similarity 0.88)
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.
