File size: 9,786 Bytes
391a57b
1d3bc55
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
391a57b
 
af103c4
391a57b
1d3bc55
 
 
 
 
 
 
 
b53e211
1d3bc55
 
 
 
 
 
391a57b
af103c4
1d3bc55
af103c4
1d3bc55
af103c4
1d3bc55
 
 
a63f065
1d3bc55
 
 
 
 
 
 
 
 
af103c4
1d3bc55
 
 
 
 
 
af103c4
1d3bc55
 
 
 
a63f065
1d3bc55
 
af103c4
1d3bc55
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
af103c4
 
 
 
 
 
1d3bc55
 
 
 
 
af103c4
 
 
 
 
 
 
 
 
1d3bc55
 
 
 
 
 
af103c4
 
 
 
 
1d3bc55
af103c4
1d3bc55
 
 
 
 
 
 
 
 
 
 
 
af103c4
1d3bc55
 
 
 
 
 
 
 
 
 
 
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
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
---
license: apache-2.0
task_categories:
  - question-answering
  - visual-question-answering
  - image-to-image
tags:
  - time series
  - time series understanding
  - time series generation
  - time series forecasting
  - time series imputation
  - multimodal
  - temporal reasoning
  - understanding-guided generation
  - timeomni-vl
language:
  - en
size_categories:
  - 10K<n<100K
---

# 🧱 TSUMM-Suite Training Set

<p align="left">
  <a href="https://arxiv.org/abs/2602.17149">
    <img
      src="https://img.shields.io/badge/TimeOmni--VL-Paper-red?logo=arxiv&logoColor=red"
      style="display: inline-block; vertical-align: middle;"
      alt="TimeOmni-VL Paper on arXiv"
    />
  </a>
  <a href="https://huggingface.co/TimeOmni-VL/TimeOmni-VL">
    <img
      src="https://img.shields.io/badge/TimeOmni--VL-Model-yellow?logo=huggingface&logoColor=white"
      style="display: inline-block; vertical-align: middle;"
      alt="TimeOmni-VL Model on Hugging Face"
    />
  </a>

  <a href="https://huggingface.co/datasets/TimeOmni-VL/TSUMM-Suite_Training">
    <img
      src="https://img.shields.io/badge/TSUMM--Suite-Dataset-orange?logo=huggingface&logoColor=white"
      style="display: inline-block; vertical-align: middle;"
      alt="TSUMM-Suite Dataset on Hugging Face"
    />
  </a>

  <a href="https://huggingface.co/spaces/anton-hugging/TimeOmni-VL">
    <img
      src="https://img.shields.io/badge/TimeOmni--VL-Demo-blue?logo=huggingface&logoColor=white"
      style="display: inline-block; vertical-align: middle;"
      alt="TimeOmni-VL Demo on Hugging Face Spaces"
    />
  </a>

  <a href="https://github.com/AntonGuan/TimeOmni-VL" target="_blank" style="margin: 2px;">
    <img
      src="https://img.shields.io/badge/TimeOmni--VL-Code-536af5?logo=github&logoColor=white"
      style="display: inline-block; vertical-align: middle;"
      alt="TimeOmni-VL Code on GitHub"
    />
  </a>
</p>

> **This repository provides the training set of TSUMM-Suite, a multimodal dataset for unified time series understanding and generation.** It contains forecasting and imputation tasks for TS-image generation. By aligning understanding tasks with generation tasks, TSUMM-Suite enables temporal understanding to improve time series generation.

## 🎨 Task Illustration

<div align="center">
<img src="https://raw.githubusercontent.com/AntonGuan/TimeOmni-VL/main/figs/task.png" width="100%"/>
</div>

<p><strong>TSUMM-Suite covers two task families:</strong></p>
<div style="margin-left:18px">

<p><strong>Generation tasks</strong></p>
<ul>
  <li><strong>Forecasting.</strong> Complete the masked future region of a TS-image and decode the generated image region back into future numerical values.</li>
  <li><strong>Imputation.</strong> Complete masked internal regions of a TS-image and decode the generated image region back into missing numerical values.</li>
</ul>

<p><strong>Understanding tasks</strong></p>
<ul>
  <li><strong>QA1: Variable Counting.</strong> Count the number of variables encoded in a TS-image.</li>
  <li><strong>QA2: Variable Y-Range.</strong> Locate the vertical range of a specific variable.</li>
  <li><strong>QA3: Cycle Bounding Box.</strong> Locate the spatial region of a target cycle.</li>
  <li><strong>QA4: Mean Comparison.</strong> Compare average values across cycles.</li>
  <li><strong>QA5: Anomaly Detection.</strong> Identify and locate anomalous regions.</li>
  <li><strong>QA6: Trend Analysis.</strong> Analyze temporal patterns within a specified region.</li>
</ul>

</div>

## 📊 Dataset Information

### 1. Dataset Components

<table style="width:100%;border-collapse:collapse;font-size:13px">
  <thead>
    <tr>
      <th style="padding:10px 7px;text-align:center;vertical-align:middle;font-weight:700;border-bottom:2px solid #2563eb;color:#2563eb">Component</th>
      <th style="padding:10px 7px;text-align:center;vertical-align:middle;font-weight:700;border-bottom:2px solid #2563eb;color:#2563eb">Task Family</th>
      <th style="padding:10px 7px;text-align:center;vertical-align:middle;font-weight:700;border-bottom:2px solid #2563eb;color:#2563eb">Size</th>
      <th style="padding:10px 7px;text-align:center;vertical-align:middle;font-weight:700;border-bottom:2px solid #2563eb;color:#2563eb">Description</th>
    </tr>
  </thead>
  <tbody>
    <tr>
      <td style="padding:8px 7px;text-align:center;vertical-align:middle;border-bottom:1px solid rgba(128,128,128,0.15);color:#2563eb">Forecasting</td>
      <td style="padding:8px 7px;text-align:center;vertical-align:middle;border-bottom:1px solid rgba(128,128,128,0.15)"><span style="display:inline-block;padding:3px 9px;border-radius:999px;background:rgba(37,99,235,0.10);color:#2563eb;font-weight:600">Generation</span></td>
      <td style="padding:8px 7px;text-align:center;vertical-align:middle;border-bottom:1px solid rgba(128,128,128,0.15);font-weight:700">40,000</td>
      <td style="padding:8px 7px;text-align:center;vertical-align:middle;border-bottom:1px solid rgba(128,128,128,0.15)">TS-image completion samples for future sequence generation</td>
    </tr>
    <tr>
      <td style="padding:8px 7px;text-align:center;vertical-align:middle;border-bottom:1px solid rgba(128,128,128,0.15);color:#2563eb">Imputation</td>
      <td style="padding:8px 7px;text-align:center;vertical-align:middle;border-bottom:1px solid rgba(128,128,128,0.15)"><span style="display:inline-block;padding:3px 9px;border-radius:999px;background:rgba(37,99,235,0.10);color:#2563eb;font-weight:600">Generation</span></td>
      <td style="padding:8px 7px;text-align:center;vertical-align:middle;border-bottom:1px solid rgba(128,128,128,0.15);font-weight:700">40,000</td>
      <td style="padding:8px 7px;text-align:center;vertical-align:middle;border-bottom:1px solid rgba(128,128,128,0.15)">TS-image completion samples for missing-value reconstruction</td>
    </tr>
    <tr>
      <td style="padding:8px 7px;text-align:center;vertical-align:middle;border-bottom:1px solid rgba(128,128,128,0.15);color:#2563eb">TS-image QA</td>
      <td style="padding:8px 7px;text-align:center;vertical-align:middle;border-bottom:1px solid rgba(128,128,128,0.15)"><span style="display:inline-block;padding:3px 9px;border-radius:999px;background:rgba(37,99,235,0.10);color:#2563eb;font-weight:600">Understanding</span></td>
      <td style="padding:8px 7px;text-align:center;vertical-align:middle;border-bottom:1px solid rgba(128,128,128,0.15);font-weight:700">9,409</td>
      <td style="padding:8px 7px;text-align:center;vertical-align:middle;border-bottom:1px solid rgba(128,128,128,0.15)">CoT-guided QA pairs for layout-level and signal-level TS-image understanding</td>
    </tr>
    <tr>
      <td style="padding:8px 7px;text-align:center;vertical-align:middle;border-bottom:1px solid rgba(128,128,128,0.15);color:#2563eb">TSR-Suite</td>
      <td style="padding:8px 7px;text-align:center;vertical-align:middle;border-bottom:1px solid rgba(128,128,128,0.15)"><span style="display:inline-block;padding:3px 9px;border-radius:999px;background:rgba(37,99,235,0.10);color:#2563eb;font-weight:600">Reasoning</span></td>
      <td style="padding:8px 7px;text-align:center;vertical-align:middle;border-bottom:1px solid rgba(128,128,128,0.15);font-weight:700">2,339</td>
      <td style="padding:8px 7px;text-align:center;vertical-align:middle;border-bottom:1px solid rgba(128,128,128,0.15)">CoT-guided temporal reasoning samples used to strengthen temporal reasoning</td>
    </tr>
  </tbody>
</table>

<p style="margin-top:13px;font-size:12px;opacity:0.85">
* Note: This release open-sources the forecasting and imputation generation data only. We do not open-source the TS-image QA (Understanding) and TSR-Suite (Reasoning) training data.
</p>

All generation samples are stored in a single `train` split (80,000 rows total). Each TS-image has resolution **896×896**. In the paper, the main TimeOmni-VL experiments use **5k samples per generation task** for training.

### 2. Data Fields

Generation samples contain fields such as:

- `sample_id`: unique sample identifier
- `task`: `forecasting` or `imputation`
- `freq`: time series frequency (e.g., `H`, `D`, `5T`)
- `nvars`: number of variables
- `context_len`: context length
- `pred_len`: prediction or imputation length
- `periodicity`: periodicity of the series
- `input_image`: masked TS-image input (896×896)
- `target_image`: ground-truth completed TS-image (896×896)
- `metadata_json`: JSON string with normalization, patch layout, mask ratio, and related metadata

## 🚀 Usage

```python
from datasets import load_dataset

dataset = load_dataset("TimeOmni-VL/TSUMM-Suite_Training")
train = dataset["train"]

forecasting = train.filter(lambda x: x["task"] == "forecasting")
imputation = train.filter(lambda x: x["task"] == "imputation")

print(f"Total: {len(train)}, Forecasting: {len(forecasting)}, Imputation: {len(imputation)}")
```

## 📐 Evaluation

**For forecasting and imputation**, we report normalized Mean Absolute Scaled Error (nMASE). Lower values indicate better numerical generation quality.

**For TS-image understanding**, we report normalized task-specific scores in the range of `[0, 1]`. Higher values indicate better understanding performance.

**For TSR-Suite reasoning tasks**, we report Accuracy (ACC) for text-output tasks and Mean Absolute Error (MAE) for sequence-output tasks.

## License

TSUMM-Suite is released under the Apache 2.0 license.

## ✍️ Citation

```bibtex
@article{guan2026timeomni,
  title={TimeOmni-VL: Unified Models for Time Series Understanding and Generation},
  author={Guan, Tong and Pan, Sheng and Barthelemy, Johan and Li, Zhao and Cai, Yujun and Alippi, Cesare and Jin, Ming and Pan, Shirui},
  journal={arXiv preprint arXiv:2602.17149},
  year={2026}
}
```