The dataset viewer is not available for this split.
Error code: TooBigContentError
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
MTBench: A Multimodal Time Series Benchmark
MTBench (Huggingface, Github, Arxiv) is a suite of multimodal datasets for evaluating large language models (LLMs) in temporal and cross-modal reasoning tasks across finance and weather domains.
Each benchmark instance aligns high-resolution time series (e.g., stock prices, weather data) with textual context (e.g., news articles, QA prompts), enabling research into temporally grounded and multimodal understanding.
๐ฆ Stock Time-series
We provide high-resolution time series data for 2,993 U.S. stocks spanning a 10-year period (2013โ2023). The data is recorded at 5-minute intervals, offering fine-grained temporal resolution for modeling and analysis.
Each stock record includes the following attributes:
- Open, High, Low, Close prices (OHLC)
- Volume of shares traded
- VWAP (Volume-Weighted Average Price)
- Number of Transactions
This dataset enables detailed financial forecasting, event correlation, and temporal pattern analysis.
๐ฆ Other MTBench Datasets
๐น Finance Domain
MTBench_finance_news
20,000 articles with URL, timestamp, context, and labelsMTBench_finance_stock
Time series of 2,993 stocks (2013โ2023)MTBench_finance_aligned_pairs_short
2,000 newsโseries pairs- Input: 7 days @ 5-min
- Output: 1 day @ 5-min
MTBench_finance_aligned_pairs_long
2,000 newsโseries pairs- Input: 30 days @ 1-hour
- Output: 7 days @ 1-hour
MTBench_finance_QA_short
490 multiple-choice QA pairs- Input: 7 days @ 5-min
- Output: 1 day @ 5-min
MTBench_finance_QA_long
490 multiple-choice QA pairs- Input: 30 days @ 1-hour
- Output: 7 days @ 1-hour
๐น Weather Domain
MTBench_weather_news
Regional weather event descriptionsMTBench_weather_temperature
Meteorological time series from 50 U.S. stationsMTBench_weather_aligned_pairs_short
Short-range aligned weather textโseries pairsMTBench_weather_aligned_pairs_long
Long-range aligned weather textโseries pairsMTBench_weather_QA_short
Short-horizon QA with aligned weather dataMTBench_weather_QA_long
Long-horizon QA for temporal and contextual reasoning
๐ง Supported Tasks
MTBench supports a wide range of multimodal and temporal reasoning tasks, including:
- ๐ News-aware time series forecasting
- ๐ Event-driven trend analysis
- โ Multimodal question answering (QA)
- ๐ Text-to-series correlation analysis
- ๐งฉ Causal inference in financial and meteorological systems
๐ Citation
If you use MTBench in your work, please cite:
@article{chen2025mtbench,
title={MTBench: A Multimodal Time Series Benchmark for Temporal Reasoning and Question Answering},
author={Chen, Jialin and Feng, Aosong and Zhao, Ziyu and Garza, Juan and Nurbek, Gaukhar and Qin, Cheng and Maatouk, Ali and Tassiulas, Leandros and Gao, Yifeng and Ying, Rex},
journal={arXiv preprint arXiv:2503.16858},
year={2025}
}
- Downloads last month
- 3,031