File size: 4,654 Bytes
aaa2c14
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7c112fc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
---
dataset_info:
  features:
  - name: DATE
    sequence: string
  - name: temperature
    sequence: float32
  - name: dew_point_temperature
    sequence: float32
  - name: relative_humidity
    sequence: float32
  - name: station_level_pressure
    sequence: float32
  - name: sea_level_pressure
    sequence: float32
  - name: wind_speed
    sequence: float32
  - name: wind_direction
    sequence: float32
  - name: visibility
    sequence: float32
  - name: altimeter
    sequence: float32
  - name: precipitation_3_hour
    sequence: float32
  - name: precipitation_6_hour
    sequence: float32
  - name: precipitation_9_hour
    sequence: float32
  - name: precipitation_12_hour
    sequence: float32
  - name: precipitation_15_hour
    sequence: float32
  - name: precipitation_18_hour
    sequence: float32
  - name: precipitation_21_hour
    sequence: float32
  - name: precipitation_24_hour
    sequence: float32
  splits:
  - name: train
    num_bytes: 1173537799
    num_examples: 49
  download_size: 168154556
  dataset_size: 1173537799
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---
# MTBench: A Multimodal Time Series Benchmark


**MTBench** ([Huggingface](https://huggingface.co/collections/afeng/mtbench-682577471b93095c0613bbaa), [Github](https://github.com/Graph-and-Geometric-Learning/MTBench), [Arxiv](https://arxiv.org/pdf/2503.16858)) 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.

## 📦 MTBench Datasets

### 🔹 Finance Domain

- [`MTBench_finance_news`](https://huggingface.co/datasets/afeng/MTBench_finance_news)  
  20,000 articles with URL, timestamp, context, and labels

- [`MTBench_finance_stock`](https://huggingface.co/datasets/afeng/MTBench_finance_stock)  
  Time series of 2,993 stocks (2013–2023)

- [`MTBench_finance_aligned_pairs_short`](https://huggingface.co/datasets/afeng/MTBench_finance_aligned_pairs_short)  
  2,000 news–series pairs  
  - Input: 7 days @ 5-min  
  - Output: 1 day @ 5-min

- [`MTBench_finance_aligned_pairs_long`](https://huggingface.co/datasets/afeng/MTBench_finance_aligned_pairs_long)  
  2,000 news–series pairs  
  - Input: 30 days @ 1-hour  
  - Output: 7 days @ 1-hour

- [`MTBench_finance_QA_short`](https://huggingface.co/datasets/afeng/MTBench_finance_QA_short)  
  490 multiple-choice QA pairs  
  - Input: 7 days @ 5-min  
  - Output: 1 day @ 5-min

- [`MTBench_finance_QA_long`](https://huggingface.co/datasets/afeng/MTBench_finance_QA_long)  
  490 multiple-choice QA pairs  
  - Input: 30 days @ 1-hour  
  - Output: 7 days @ 1-hour

### 🔹 Weather Domain

- [`MTBench_weather_news`](https://huggingface.co/datasets/afeng/MTBench_weather_news)  
  Regional weather event descriptions

- [`MTBench_weather_temperature`](https://huggingface.co/datasets/afeng/MTBench_weather_temperature)  
  Meteorological time series from 50 U.S. stations

- [`MTBench_weather_aligned_pairs_short`](https://huggingface.co/datasets/afeng/MTBench_weather_aligned_pairs_short)  
  Short-range aligned weather text–series pairs

- [`MTBench_weather_aligned_pairs_long`](https://huggingface.co/datasets/afeng/MTBench_weather_aligned_pairs_long)  
  Long-range aligned weather text–series pairs

- [`MTBench_weather_QA_short`](https://huggingface.co/datasets/afeng/MTBench_weather_QA_short)  
  Short-horizon QA with aligned weather data

- [`MTBench_weather_QA_long`](https://huggingface.co/datasets/afeng/MTBench_weather_QA_long)  
  Long-horizon QA for temporal and contextual reasoning



## 🧠 Supported Tasks

MTBench supports a wide range of multimodal and temporal reasoning tasks, including:

- 📈 **News-aware time series forecasting**
- 📊 **Event-driven trend analysis**
-**Multimodal question answering (QA)**
- 🔄 **Text-to-series correlation analysis**
- 🧩 **Causal inference in financial and meteorological systems**



## 📄 Citation

If you use MTBench in your work, please cite:

```bibtex
@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}
}