Research map / Time series / Causal and generative time series
A Theoretical Analysis of Detecting Large Model-Generated Time Series
Junji Hou, Junzhou Zhao, Shuo Zhang, Pinghui Wang
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 |
|---|---|---|---|
| use entropy of internal distribution (UCE-Entropy) vs best text baseline DetectLLM-LLR | detection · In-distribution (12 datasets) | AUROC: 0.815 -> 0.855 | Improved |
| UCE-Entropy vs DetectLLM-LLR in zero-shot | detection · Zero-shot (20 datasets) | AUROC: 0.705 -> 0.731 | Improved |
| UCE-Entropy vs DetectLLM-LLR TPR at 1% FPR in zero-shot | detection · Zero-shot (20 datasets) | TPR at 1% FPR: 0.233 -> 0.286 | Improved |
| use Variance instead of Entropy as uncertainty metric | UCE · Zero-shot (20 datasets) | AUROC: 0.731 -> 0.634 | Hurt |
| use Max-Probability instead of Entropy | UCE · Zero-shot (20 datasets) | TPR at 1% FPR: TPR 0.286 -> 0.306 | No clear effect |
| apply UCE-Entropy to Time-MoE generated series | UCE · Time-MoE, 9 datasets | AUROC: AUROC 0.829 -> 0.957 (H = 96 to H = 336) | Improved |
| use Max-Probability on Timer-generated series | UCE · Timer, 9 datasets | AUROC: 0.833 -> 0.556 (Entropy vs MaxProb, H = 96) | Hurt |
| text-based Binocular detector applied to time series | Binocular · In-distribution (12 datasets) | AUROC: AUROC 0.523 in-distribution | Hurt |
Closest papers by meaning
- Can LLMs Understand Time Series Anomalies? (ICLR 2025 · similarity 0.89)
- TSGM: A Flexible Framework for Generative Modeling of Synthetic Time Series (NeurIPS 2024 · similarity 0.88)
- Time Series Generation Under Data Scarcity: A Unified Generative Modeling Approach (NeurIPS 2025 · similarity 0.87)
- Synthetic Series-Symbol Data Generation for Time Series Foundation Models (NeurIPS 2025 · similarity 0.87)
- Timer: Generative Pre-trained Transformers Are Large Time Series Models (ICML 2024 · similarity 0.87)
- VerbalTS: Generating Time Series from Texts (ICML 2025 · similarity 0.86)
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.
