TimeSeriesExam: leaderboard
Conceptual time-series understanding on TimeSeriesExam (2026): 763 multiple-choice questions on synthetic series built from over 100 expert-curated templates in five categories (pattern recognition, noise understanding, similarity analysis, anomaly detection, causality), with items refined by item response theory. Overall accuracy in percent; higher is better.
Metric: Overall accuracy (%, times 100) on the 763 questions of TimeSeriesExam, 763 multiple-choice questions on synthetic series generated from over 100 expert-curated templates and refined over rounds with item response theory to separate candidate models; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 4 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 73 | #751 |
| 2 | Gemini 2.5 Pro | 71 | #276 |
| 3 | Gemma 3 27B (IT) | 59 | #1324 |
| 4 | Qwen 2.5 VL 72B Instruct | 44 | #841 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=timeseriesexam · How It Works · Data refreshed daily, snapshot 2026-11-21.