Research map / Time series / Causal and generative time series
ConTSG-Bench: A Unified Benchmark for Conditional Time Series Generation
Shaocheng Lan, Shuqi Gu, Zhangzhi Xiong, Kan Ren
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 |
|---|---|---|---|
| evaluate condition adherence separately from generation fidelity | ConTSG-Bench evaluation · overall ranking of 11 models | average rank | Improved |
| use text conditioning instead of attribute or label conditioning | conditional time series generators · ConTSG-Bench overall ranking | average rank | Mixed |
| use conceptual instead of morphological conditions (ablation) | conditional generators · PTB-XL | DTW / CRPS | No clear effect |
| use conceptual instead of morphological conditions (ablation) | conditional generators · Weather | DTW / CRPS | Improved |
| text-conditioned generators versus naive retrieval baseline on segment-level patterns | text-conditioned generators · Synth-U shapelet joint accuracy | joint accuracy | Mixed |
| increase candidate pool size for segment retrieval (ablation) | text-conditioned generators · TelecomTS-Segment | Acc@1 | Hurt |
| segment-level temporal order control | all evaluated generators · TelecomTS-Segment | temporal order accuracy | No clear effect |
| move test conditions from near to far from training attribute combinations (ablation) | most generators · compositional generalization (Hamming distance kNN) | normalized retrieval accuracy | Hurt |
| stronger condition adherence (e.g. VerbalTS) under novel combinations | VerbalTS · compositional generalization | normalized retrieval accuracy | Mixed |
| train downstream classifier on generated instead of real data | conditional generators · downstream attribute classification | drop rate | Mixed |
Closest papers by meaning
- Time Series Generation Under Data Scarcity: A Unified Generative Modeling Approach (NeurIPS 2025 · similarity 0.89)
- FIDE: Frequency-Inflated Conditional Diffusion Model for Extreme-Aware Time Series Generation (NeurIPS 2024 · similarity 0.87)
- Time Weaver: A Conditional Time Series Generation Model (ICML 2024 · similarity 0.87)
- Tractable Transformers for Flexible Conditional Generation (ICML 2025 · similarity 0.86)
- BRIDGE: Bootstrapping Text to Control Time-Series Generation via Multi-Agent Iterative Optimization and Diffusion Modeling (ICML 2025 · similarity 0.86)
- VerbalTS: Generating Time Series from Texts (ICML 2025 · similarity 0.85)
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.
