Dataset Preview
Duplicate
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed because of a cast error
Error code:   DatasetGenerationCastError
Exception:    DatasetGenerationCastError
Message:      An error occurred while generating the dataset

All the data files must have the same columns, but at some point there are 4 new columns ({'prompt_3', 'option', 'truth_1', 'truth'}) and 5 missing columns ({'timeseries_2', 'timeseries_1', 'timeseries_3', 'c_meta', 'timeseries_0'}).

This happened while the json dataset builder was generating data using

hf://datasets/November-Rain/HiTSR/Train/l2_train.json (at revision 566be99bf403051df326eeb883d4e3888cbe508b), [/tmp/hf-datasets-cache/medium/datasets/42635665187954-config-parquet-and-info-November-Rain-HiTSR-4ef1d656/hub/datasets--November-Rain--HiTSR/snapshots/566be99bf403051df326eeb883d4e3888cbe508b/Train/l1_train.json (origin=hf://datasets/November-Rain/HiTSR@566be99bf403051df326eeb883d4e3888cbe508b/Train/l1_train.json), /tmp/hf-datasets-cache/medium/datasets/42635665187954-config-parquet-and-info-November-Rain-HiTSR-4ef1d656/hub/datasets--November-Rain--HiTSR/snapshots/566be99bf403051df326eeb883d4e3888cbe508b/Train/l2_train.json (origin=hf://datasets/November-Rain/HiTSR@566be99bf403051df326eeb883d4e3888cbe508b/Train/l2_train.json), /tmp/hf-datasets-cache/medium/datasets/42635665187954-config-parquet-and-info-November-Rain-HiTSR-4ef1d656/hub/datasets--November-Rain--HiTSR/snapshots/566be99bf403051df326eeb883d4e3888cbe508b/Train/l3_train.json (origin=hf://datasets/November-Rain/HiTSR@566be99bf403051df326eeb883d4e3888cbe508b/Train/l3_train.json)], ['hf://datasets/November-Rain/HiTSR@566be99bf403051df326eeb883d4e3888cbe508b/Train/l1_train.json', 'hf://datasets/November-Rain/HiTSR@566be99bf403051df326eeb883d4e3888cbe508b/Train/l2_train.json', 'hf://datasets/November-Rain/HiTSR@566be99bf403051df326eeb883d4e3888cbe508b/Train/l3_train.json']

Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1800, in _prepare_split_single
                  writer.write_table(table)
                File "/usr/local/lib/python3.12/site-packages/datasets/arrow_writer.py", line 765, in write_table
                  self._write_table(pa_table, writer_batch_size=writer_batch_size)
                File "/usr/local/lib/python3.12/site-packages/datasets/arrow_writer.py", line 773, in _write_table
                  pa_table = table_cast(pa_table, self._schema)
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2321, in table_cast
                  return cast_table_to_schema(table, schema)
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2249, in cast_table_to_schema
                  raise CastError(
              datasets.table.CastError: Couldn't cast
              id: int64
              truth: string
              option: string
              2img_prompt: string
              prompt_1: string
              prompt_2: string
              prompt_3: string
              answer_1: string
              answer_2: string
              timeseries: list<item: list<item: double>>
                child 0, item: list<item: double>
                    child 0, item: double
              truth_1: string
              to
              {'id': Value('int64'), 'timeseries': List(List(Value('float64'))), '2img_prompt': Value('string'), 'prompt_1': Value('string'), 'prompt_2': Value('string'), 'answer_1': Value('string'), 'answer_2': Value('string'), 'timeseries_0': List(List(Value('float64'))), 'timeseries_1': List(List(Value('float64'))), 'timeseries_2': List(List(Value('float64'))), 'c_meta': {'mode': Value('string'), 'winner_series': Value('int64'), 'winner_index': Value('int64'), 'winner_value': Value('float64'), 'T': Value('int64'), 'D': Value('int64'), 'selected_source_indices': List(Value('int64'))}, 'timeseries_3': List(List(Value('float64')))}
              because column names don't match
              
              During handling of the above exception, another exception occurred:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1347, in compute_config_parquet_and_info_response
                  parquet_operations = convert_to_parquet(builder)
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 980, in convert_to_parquet
                  builder.download_and_prepare(
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 882, in download_and_prepare
                  self._download_and_prepare(
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 943, in _download_and_prepare
                  self._prepare_split(split_generator, **prepare_split_kwargs)
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1646, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1802, in _prepare_split_single
                  raise DatasetGenerationCastError.from_cast_error(
              datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
              
              All the data files must have the same columns, but at some point there are 4 new columns ({'prompt_3', 'option', 'truth_1', 'truth'}) and 5 missing columns ({'timeseries_2', 'timeseries_1', 'timeseries_3', 'c_meta', 'timeseries_0'}).
              
              This happened while the json dataset builder was generating data using
              
              hf://datasets/November-Rain/HiTSR/Train/l2_train.json (at revision 566be99bf403051df326eeb883d4e3888cbe508b), [/tmp/hf-datasets-cache/medium/datasets/42635665187954-config-parquet-and-info-November-Rain-HiTSR-4ef1d656/hub/datasets--November-Rain--HiTSR/snapshots/566be99bf403051df326eeb883d4e3888cbe508b/Train/l1_train.json (origin=hf://datasets/November-Rain/HiTSR@566be99bf403051df326eeb883d4e3888cbe508b/Train/l1_train.json), /tmp/hf-datasets-cache/medium/datasets/42635665187954-config-parquet-and-info-November-Rain-HiTSR-4ef1d656/hub/datasets--November-Rain--HiTSR/snapshots/566be99bf403051df326eeb883d4e3888cbe508b/Train/l2_train.json (origin=hf://datasets/November-Rain/HiTSR@566be99bf403051df326eeb883d4e3888cbe508b/Train/l2_train.json), /tmp/hf-datasets-cache/medium/datasets/42635665187954-config-parquet-and-info-November-Rain-HiTSR-4ef1d656/hub/datasets--November-Rain--HiTSR/snapshots/566be99bf403051df326eeb883d4e3888cbe508b/Train/l3_train.json (origin=hf://datasets/November-Rain/HiTSR@566be99bf403051df326eeb883d4e3888cbe508b/Train/l3_train.json)], ['hf://datasets/November-Rain/HiTSR@566be99bf403051df326eeb883d4e3888cbe508b/Train/l1_train.json', 'hf://datasets/November-Rain/HiTSR@566be99bf403051df326eeb883d4e3888cbe508b/Train/l2_train.json', 'hf://datasets/November-Rain/HiTSR@566be99bf403051df326eeb883d4e3888cbe508b/Train/l3_train.json']
              
              Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)

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.

id
int64
timeseries
list
2img_prompt
string
prompt_1
string
prompt_2
string
answer_1
string
answer_2
string
timeseries_0
list
timeseries_1
list
timeseries_2
list
c_meta
dict
timeseries_3
null
0
[ [ 64.054444, 76.885919, 71.03885, 87.839298, 83.633652, 82.579088, 81.167753, 77.637582, 81.230013, 87.656862, 75.966002, 90.971937, 82.398177, 88.387103, 76.450253, 90.393108, 98.772127, 99.458126, 87.450837, 87.327619, 95.181541, ...
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.You must output the max and min values in the following format: <max>[index_ma...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.
<max>[185, 102.200264]</max> <min>[305, 48.113776]</min> The max value appears first.
The maximum value is 102.200 at index 185. The minimum value is 48.114 at index 305. The minimum point appears after the maximum point.
null
null
null
null
null
1
[ [ -1.180848, -0.418106, 0.632636, 1.667433, 2.714672, 3.765568, 4.807774, 5.857745, 6.919028, 7.967587, 8.967088, 7.697964, 6.416627, 5.158746, 3.875234, 2.605336, 1.337367, 0.066471, -1.19404, -2.477862, -3.423366, -3.51733, ...
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.You must output the max and min values in the following format: <max>[index_ma...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.
<max>[10, 8.967088]</max> <min>[92, -7.377136]</min> The max value appears first.
The maximum value is 8.967 at index 10. The minimum value is -7.377 at index 92. The minimum point appears after the maximum point.
null
null
null
null
null
2
null
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
Among all provided series, identify the one with the highest peak. Give the peak's first occurrence index and its precise numerical value. The start index is set to 0.
Given multiple time series, compare the maximum values across the given time series and determine which series has the highest maximum. You must use the FIRST occurrence index of that maximum and output using the required tags. The start index is set to 0. You MUST output exactly in this format: <answer>[index, value]...
Time Series 2 has the highest maximum value among the series. Its exact maximum value is 805.323135.
<answer>[12, 805.323135]</answer> <series>2</series> Time Series 2 has the highest maximum value among the series. Its exact maximum value is 805.323135.
[ [ -0.074951, -0.074879, -0.074911, -0.074699, -0.074685, -0.074598, -0.074533, -0.074432, -0.074354, -0.074306, -0.074221, -0.074196, -0.074068, -0.074057, -0.07391, -0.073817, -0.073842, -0.073739, -0.07367, -0.073662, -0.073559, ...
[ [ 568.304956, 504.294856, 487.206376, 625.53052, 542.404898, 565.293498, 680.227493, 683.083326, 624.81057, 721.555746, 764.823605, 467.437666, 805.323135, 733.224386, 663.877017, 781.494069, 688.726363, 640.518362, 648.105515, 556.7633...
[ [ -0.060559, -0.059988, -0.059436, -0.058882, -0.058331, -0.057764, -0.057224, -0.056663, -0.056123, -0.055577, -0.055021, -0.054472, -0.053921, -0.05338, -0.052828, -0.05228, -0.051738, -0.051196, -0.050668, -0.050125, -0.049587, ...
{ "mode": "max_compare", "winner_series": 2, "winner_index": 12, "winner_value": 805.323135, "T": 566, "D": 3, "selected_source_indices": [ 7216, 12411, 5057 ] }
null
3
[ [ -1202811.262015, -1221086.340211, -1247287.974982, -1339513.543627, -1250735.465231, -1254072.983246, -1308230.23613, -1307924.281135, -1259506.075446, -1218340.995657, -1265682.246312, -1373078.649457, -1300257.862794, -1311438.055195, -1366426.950815, ...
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.You must output the max and min values in the following format: <max>[index_ma...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.
<max>[154, 1879438.710185]</max> <min>[53, -4731800.28905]</min> The min value appears first.
The maximum value is 1879438.710 at index 154. The minimum value is -4731800.289 at index 53. The minimum point appears before the maximum point.
null
null
null
null
null
4
null
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
Retrieve the specific numerical values for **Series0** from index 139 to index 149 (inclusive). Organize the result as a list and do not output any other text.
Retrieve the specific numerical values for **Series0** from index 139 to index 149 (inclusive). You MUST output exactly in the following format and do not output any other text: <values>[v1, v2, ...]</values>
The values for Series0 from index 139 to 149 are: [870.035, 798.081, 760.888, 762.845, 806.423, 886.375, 995.982, 1121.519, 1250.519, 1368.030, 1459.987].
<values>[870.035, 798.081, 760.888, 762.845, 806.423, 886.375, 995.982, 1121.519, 1250.519, 1368.030, 1459.987]</values>
[ [ -79.896781, -36.126809, 42.445852, 145.628274, 262.605894, 376.5069, 475.831955, 545.361147, 578.995959, 571.587386, 525.159659, 443.385145, 338.596545, 222.976246, 111.32486, 16.059942, -48.392758, -75.142703, -61.286196, -7.612527, ...
null
null
null
null
5
[ [ -33.663198, -33.695738, -33.665551, -33.691682, -33.674046, -33.688658, -33.716262, -33.724069, -33.666255, -33.663032, -33.687828, -33.690905, -33.698606, -33.736673, -33.690239, -33.682981, -33.681405, -33.69937, -33.708713, -33.653...
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
Identify the numerical values of the starting and ending points of the given time series, and describe their magnitude relationship. The start index is set to 0.
Identify the numerical values of the starting and ending points of the given time series, and describe their magnitude relationship. The start index is set to 0.You must output the starting and ending values in the following format: <start>[start_index, start_value]</start> <end>[end_index, end_value]</end> Then state ...
The starting value is -33.663 at index 0. The ending value is -33.703 at the final index. The starting value is higher than the ending value.
<start>[0, -33.663]</start> <end>[204, -33.703]</end> The ending value is lower than the starting value.
null
null
null
null
null
6
null
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
Given multiple time series, compare the maximum values across the given time series. Which series attains the largest maximum? Report the first index at which this maximum occurs and the exact value. The start index is set to 0.
Given multiple time series, compare the maximum values across the given time series and determine which series has the highest maximum. You must use the FIRST occurrence index of that maximum and output using the required tags. The start index is set to 0. You MUST output exactly in this format: <answer>[index, value]...
Time Series 3 has the highest maximum value among the series. Its exact maximum value is 10.42577.
<answer>[70, 10.42577]</answer> <series>3</series> Time Series 3 has the highest maximum value among the series. Its exact maximum value is 10.42577.
[ [ 0.088172, 0.106459, 0.11001, 0.100163, 0.102357, 0.112371, 0.098058, 0.099195, 0.104376, 0.106646, 0.135676, 0.118342, 0.10692, 0.091273, 0.117426, 0.116494, 0.122693, 0.120382, 0.105484, 0.100269, 0.075819, 0.085758, 0.08...
[ [ -0.669786, -0.711035, -0.755693, -0.792728, -0.832759, -0.871479, -0.907443, -0.946974, -0.990187, -1.029936, -1.071362, -1.107473, -1.14528, -1.186456, -1.226759, -1.267846, -1.301149, -1.343338, -1.382554, -1.418703, -1.460878, ...
[ [ -0.185125, -0.182374, -0.178069, -0.17505, -0.171249, -0.168178, -0.164647, -0.160783, -0.157361, -0.153594, -0.15063, -0.146653, -0.144531, -0.140332, -0.137219, -0.133969, -0.129748, -0.127072, -0.123607, -0.119053, -0.117373, ...
{ "mode": "max_compare", "winner_series": 3, "winner_index": 70, "winner_value": 10.42577, "T": 498, "D": 4, "selected_source_indices": [ 12519, 4797, 1206, 19885 ] }
null
7
[ [ 0.090001, 0.089824, 0.09003, 0.090018, 0.089961, 0.090002, 0.090005, 0.090065, 0.090041, 0.090041, 0.090065, 0.090025, 0.09021, 0.090078, 0.090135, 0.09019, 0.090135, 0.090321, 0.090283, 0.090359, 0.0903, 0.09036, 0.090582...
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.You must output the max and min values in the following format: <max>[index_ma...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.
<max>[248, 0.997441]</max> <min>[1, 0.089824]</min> The min value appears first.
The maximum value is 0.997 at index 248. The minimum value is 0.090 at index 1. The minimum point appears before the maximum point.
null
null
null
null
null
8
[ [ -429.790708, -401.97241, -432.669933, -416.903986, -432.711391, -363.621886, -379.316726, -441.354417, -419.13221, -405.082047, -377.524787, -486.865808, -378.300846, -484.258988, -609.108145, -520.264434, -470.606102, -488.664775, -536.7...
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.You must output the max and min values in the following format: <max>[index_ma...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.
<max>[451, 811.453221]</max> <min>[110, -3189.60808]</min> The min value appears first.
The maximum value is 811.453 at index 451. The minimum value is -3189.608 at index 110. The minimum point appears before the maximum point.
null
null
null
null
null
9
null
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
Given multiple time series, compare the maximum values across the given time series. Which series attains the largest maximum? Report the first index at which this maximum occurs and the exact value. The start index is set to 0.
Given multiple time series, compare the maximum values across the given time series and determine which series has the highest maximum. You must use the FIRST occurrence index of that maximum and output using the required tags. The start index is set to 0. You MUST output exactly in this format: <answer>[index, value]...
Time Series 1 has the highest maximum value among the series. Its exact maximum value is 5602.714774.
<answer>[299, 5602.714774]</answer> <series>1</series> Time Series 1 has the highest maximum value among the series. Its exact maximum value is 5602.714774.
[ [ -743.533575, -731.471389, -719.344083, -707.293707, -695.263268, -683.17453, -671.113656, -659.162745, -647.267229, -635.490721, -623.569766, -611.733192, -599.841747, -588.194615, -576.520288, -564.76078, -553.134631, -541.759793, -530.0...
[ [ 0.101136, 0.067142, 0.075367, 0.081368, 0.05356, 0.075697, 0.073304, 0.064455, 0.057541, 0.045579, 0.060352, 0.054889, 0.082553, 0.052769, 0.039695, 0.046646, 0.043527, 0.039104, 0.016431, 0.044192, 0.024377, 0.033367, 0.0...
null
{ "mode": "max_compare", "winner_series": 1, "winner_index": 299, "winner_value": 5602.714774, "T": 450, "D": 2, "selected_source_indices": [ 3515, 4832 ] }
null
10
[ [ 1.019859, 0.927893, 1.113869, 0.981349, 1.049092, 1.080434, 1.05953, 0.905448, 0.856021, 0.854299, 0.977036, 0.911956, 0.791809, 0.726486, 0.799842, 0.789532, 0.739716, 0.566271, 0.765328, 0.760311, 0.679951, 0.663268, 0.6...
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.You must output the max and min values in the following format: <max>[index_ma...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.
<max>[584, 1.586936]</max> <min>[337, -0.274649]</min> The min value appears first.
The maximum value is 1.587 at index 584. The minimum value is -0.275 at index 337. The minimum point appears before the maximum point.
null
null
null
null
null
11
null
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
At the first timestep (index 0), which series has the largest initial value? Provide the exact starting value and confirm the index. The start index is set to 0.
Given multiple time series, compare the starting values (index 0) across the given time series and determine which series starts highest. Output using the required tags. The start index is set to 0. You MUST output exactly in this format: <answer>[index, value]</answer> <series>k</series> where k is the 1-based series...
Time Series 2 has the highest starting value. Its starting value is -0.039989.
<answer>[0, -0.039989]</answer> <series>2</series> Time Series 2 has the highest starting value. Its starting value is -0.039989.
[ [ -2.209464, -2.20184, -2.195221, -2.189046, -2.182938, -2.176357, -2.170696, -2.162472, -2.156142, -2.147939, -2.143139, -2.135506, -2.129322, -2.122202, -2.113935, -2.10818, -2.101602, -2.094759, -2.087296, -2.080937, -2.07287, ...
[ [ -0.039989, -0.039172, -0.038335, -0.037513, -0.036695, -0.035871, -0.035063, -0.034217, -0.033409, -0.032584, -0.031773, -0.030955, -0.030148, -0.02933, -0.028522, -0.02772, -0.02691, -0.026103, -0.025286, -0.024492, -0.023699, ...
null
{ "mode": "start_compare", "winner_series": 2, "winner_index": 0, "winner_value": -0.039989, "T": 570, "D": 2, "selected_source_indices": [ 18020, 18332 ] }
null
12
null
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
Across all series, determine which one dips to the smallest value. Report the earliest index where that minimum is reached, and give the exact minimum value. The start index is set to 0.
Given multiple time series, compare the minimum values across the given time series and determine which series has the lowest minimum. You must use the FIRST occurrence index of that minimum and output using the required tags. The start index is set to 0. You MUST output exactly in this format: <answer>[index, value]<...
Time Series 1 has the lowest minimum value among the series. Its exact minimum value is -1027.982805.
<answer>[64, -1027.982805]</answer> <series>1</series> Time Series 1 has the lowest minimum value among the series. Its exact minimum value is -1027.982805.
[ [ -1009.907475, -929.061389, -848.793665, -770.267978, -692.020298, -616.137401, -541.760153, -469.092891, -398.481708, -328.770221, -262.752789, -199.049409, -137.548859, -78.855456, -23.355104, 29.185329, 78.852539, 124.945943, 167.221464...
[ [ 2.219913, 2.313479, 2.44234, 2.556581, 2.680105, 2.802904, 2.908029, 3.024419, 3.086082, 3.155536, 3.185706, 3.208591, 3.195667, 3.142135, 3.066193, 2.931799, 2.781011, 2.598269, 2.35605, 2.104559, 1.787001, 1.446265, 1.07...
null
{ "mode": "min_compare", "winner_series": 1, "winner_index": 64, "winner_value": -1027.982805, "T": 92, "D": 2, "selected_source_indices": [ 13851, 11257 ] }
null
13
[ [ 0.90564, 0.89036, 0.873777, 0.858229, 0.84223, 0.825451, 0.810984, 0.795335, 0.779637, 0.764846, 0.749365, 0.734312, 0.719223, 0.703714, 0.689109, 0.67513, 0.658871, 0.646584, 0.630274, 0.617002, 0.602555, 0.589175, 0.5746...
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.You must output the max and min values in the following format: <max>[index_ma...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.
<max>[626, 1.625216]</max> <min>[213, -0.623143]</min> The min value appears first.
The maximum value is 1.625 at index 626. The minimum value is -0.623 at index 213. The minimum point appears before the maximum point.
null
null
null
null
null
14
[ [ -0.586193, -0.557178, -0.458659, -0.323216, -0.183361, -0.068386, 0.010389, 0.055507, 0.08382, 0.110554, 0.144307, 0.181867, 0.207868, 0.204316, 0.162882, 0.083993, -0.01114, -0.099686, -0.154867, -0.165047, -0.136324, -0.0939...
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.You must output the max and min values in the following format: <max>[index_ma...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.
<max>[328, 0.694694]</max> <min>[172, -1.26163]</min> The min value appears first.
The maximum value is 0.695 at index 328. The minimum value is -1.262 at index 172. The minimum point appears before the maximum point.
null
null
null
null
null
15
null
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
At the last timestep of each series, identify which series has the largest ending value. Return the corresponding index and exact value, and the series number. The start index is set to 0.
Given multiple time series, compare the ending values (final index) across the given time series and determine which series ends highest. Output using the required tags. The start index is set to 0. You MUST output exactly in this format: <answer>[index, value]</answer> <series>k</series> where k is the 1-based series...
Time Series 3 has the highest ending value. Its ending value is 0.48.
<answer>[270, 0.48]</answer> <series>3</series> Time Series 3 has the highest ending value. Its ending value is 0.48.
[ [ -0.107606, -0.069462, -0.026362, -0.209734, -0.119486, -0.046454, -0.1559, -0.123819, -0.009722, -0.148707, 0.000393, -0.0049, -0.134888, 0.015808, 0.055794, 0.141961, 0.008071, 0.013086, -0.059779, -0.020396, 0.054569, -0.107...
[ [ 0.209716, 0.209228, 0.207704, 0.206693, 0.205636, 0.204836, 0.203698, 0.202125, 0.201304, 0.199815, 0.198534, 0.197406, 0.19652, 0.195967, 0.193917, 0.192717, 0.191646, 0.190523, 0.190264, 0.188235, 0.18704, 0.186768, 0.18...
[ [ 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, 0.48, ...
{ "mode": "end_compare", "winner_series": 3, "winner_index": 270, "winner_value": 0.48, "T": 271, "D": 3, "selected_source_indices": [ 16753, 13973, 15748 ] }
null
16
null
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
At the first timestep (index 0), which series has the largest initial value? Provide the exact starting value and confirm the index. The start index is set to 0.
Given multiple time series, compare the starting values (index 0) across the given time series and determine which series starts highest. Output using the required tags. The start index is set to 0. You MUST output exactly in this format: <answer>[index, value]</answer> <series>k</series> where k is the 1-based series...
Time Series 2 has the highest starting value. Its starting value is 3.050414.
<answer>[0, 3.050414]</answer> <series>2</series> Time Series 2 has the highest starting value. Its starting value is 3.050414.
[ [ -76.574799, -76.515221, -76.539847, -76.52288, -76.573396, -76.590199, -76.589065, -76.71268, -76.679655, -76.73334, -76.729254, -76.900339, -76.959027, -77.003322, -77.108457, -77.156656, -77.212806, -77.333231, -77.424445, -77.49180...
[ [ 3.050414, 2.253441, 4.387753, 3.179568, 3.709837, 2.956614, 2.922659, 3.361725, 2.484401, 3.12343, 2.995293, 2.595229, 3.136073, 3.15998, 4.102986, 5.934452, 8.217874, 10.208633, 10.474942, 13.012209, 11.753724, 10.665282, ...
null
{ "mode": "start_compare", "winner_series": 2, "winner_index": 0, "winner_value": 3.050414, "T": 252, "D": 2, "selected_source_indices": [ 5724, 139 ] }
null
17
[ [ -0.00005, -0.000566, -0.001279, -0.001651, -0.002322, -0.00296, -0.003473, -0.004068, -0.004511, -0.005314, -0.005816, -0.006231, -0.006993, -0.007436, -0.008117, -0.0086, -0.0092, -0.009736, -0.010321, -0.010883, -0.011495, -...
<image><image> You must refer to the provided time series images to answer the question. The first image plots the visual trends. The second image is a **High-Density Numeric Grid** that provides precise values. **How to read the second image (Numerical Grid):** 1. **Layout**: The data is organized in a multi-column n...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.You must output the max and min values in the following format: <max>[index_ma...
Find the maximum and minimum values in the time series and report their **first occurrence indices** (i.e., if the max/min value appears multiple times, use the earliest index). Provide the exact values and then state which one appears first.
<max>[136, 0.169693]</max> <min>[447, -0.131286]</min> The max value appears first.
The maximum value is 0.170 at index 136. The minimum value is -0.131 at index 447. The minimum point appears after the maximum point.
null
null
null
null
null
End of preview.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

HiTSR Dataset

arXiv

This is the official dataset repository of the ACL 2026 Findings paper: "LLaTiSA: Towards Difficulty-Stratified Time Series Reasoning from Visual Perception to Semantics".

A comprehensive multimodal time series understanding and reasoning dataset with multiple complexity levels.

Overview

This dataset contains time series data paired with visual representations and natural language instructions for time series analysis tasks. The dataset is organized into 3 levels of complexity with corresponding train/test splits.

Dataset Statistics

  • Level 1 (Basic): Single series analysis (min/max detection, trend analysis)

    • Training samples: 54,000
    • Test samples: Multiple variants (minmax, multiseries, startend, subseries)
  • Level 2 (Intermediate): Multi-series analysis and relationships

    • Training samples: 45,632
    • Test categories: global, local, numerical
  • Level 3 (Advanced): Complex reasoning and annotations

    • Training samples: 3,515
    • Test samples: Final test set

Data Structure

Each sample contains:

{
  "id": 1,
  "timeseries": [[float_values]],
  "prompt": "Multi-modal prompt with image tags",
  "answer": "Model answer to the question",
  "2img_prompt": "Detailed instructions for image interpretation",
  "prompt_1": "Variant 1 of the question",
  "prompt_2": "Variant 2 of the question",
  "answer_1": "Answer in format 1",
  "answer_2": "Answer in format 2",
  "images": ["image_url_1", "image_url_2"]
}

Key Fields

  • id: Unique identifier for each sample
  • timeseries: Array of time series values (floats)
  • prompt: Main question/instruction with image references (e.g., <image>)
  • answer: Expected model response
  • 2img_prompt: Detailed instructions for interpreting high-density numeric grids
  • prompt_1/prompt_2: Alternative question formats
  • answer_1/answer_2: Alternative answer formats
  • images: URLs to corresponding plot and numeric grid images

Image Generation

The images field in each sample contains URLs to visual representations of the time series data. To generate these images and populate the images field:

  1. Clone the data conversion repository:

    git clone https://github.com/RainingNovember/LLaTiSA.git
    cd LLaTiSA/data_convert
    
  2. Install required dependencies:

    pip install -r requirements.txt
    
  3. Run the appropriate conversion script based on dataset level:

    For Level 1 datasets:

    python data_convert_l1.py \
      --input /path/to/level1_train.json \
      --output /path/to/level1_train_with_images.json \
      --plot_dir ./images/plots \
      --num_dir ./images/numeric \
      --plot_prefix "https://your-hosting.com/images/plots" \
      --num_prefix "https://your-hosting.com/images/numeric" \
      --sample_ratio 1.0
    

    For Level 2 datasets:

    python data_convert_l2.py \
      --input /path/to/level2_train.json \
      --output /path/to/level2_train_with_images.json \
      --plot_dir ./images/plots \
      --num_dir ./images/numeric \
      --plot_prefix "https://your-hosting.com/images/plots" \
      --num_prefix "https://your-hosting.com/images/numeric" \
      --sample_ratio 1.0
    

    For Level 3 datasets:

    python data_convert_l3.py \
      --input /path/to/level3_train.json \
      --output /path/to/level3_train_with_images.json \
      --plot_dir ./images/plots \
      --num_dir ./images/numeric \
      --plot_prefix "https://your-hosting.com/images/plots" \
      --num_prefix "https://your-hosting.com/images/numeric" \
      --sample_ratio 1.0
    
  4. Upload generated images to a hosting service (e.g., GitHub, Imgur, or cloud storage) and update the URL prefixes in the commands above.

  5. The output JSON files will have the images field populated with the correct URLs:

    {
      "images": [
        "https://your-hosting.com/images/plots/plot_1.png",
        "https://your-hosting.com/images/numeric/num_1.png"
      ]
    }
    

Notes:

  • Each script generates two types of images per sample: trend plots (plot_*.png) and high-density numeric grids (num_*.png)
  • Use --sample_ratio 1.0 to process all samples (default is 0.5)
  • The scripts automatically update the prompt and answer fields based on the dataset level requirements

File Organization

AAA-HiTSR/
β”œβ”€β”€ Train/
β”‚   β”œβ”€β”€ l1_train_llatisa.json      (Level 1: 54,000 samples)
β”‚   β”œβ”€β”€ l2_train_llatisa.json      (Level 2: 45,000+ samples)
β”‚   └── l3_train_llatisa.json      (Level 3: 3,500+ samples)
└── Test/
    β”œβ”€β”€ l1_test_startend.json
    β”œβ”€β”€ l1_test_minmax.json
    β”œβ”€β”€ l1_test_subseries.json
    β”œβ”€β”€ l1_test_multiseries.json
    β”œβ”€β”€ l2_test_global.json
    β”œβ”€β”€ l2_test_local.json
    β”œβ”€β”€ l2_test_numerical.json
    └── l3_test.json

Task Types

Level 1 - Basic Time Series Analysis

  • Finding maximum/minimum values and their indices
  • Trend detection (start/end values)
  • Subsequence identification

Level 2 - Multi-Series Analysis

  • Global patterns and relationships
  • Local anomalies and features
  • Numerical reasoning over multiple series

Level 3 - Advanced Reasoning

  • Complex queries requiring multi-step reasoning
  • Real fine-tuning (RFT) and GRPO annotations

Citation

If you use this dataset in your research, please cite:

@article{llatisa2026,
  title={LLaTiSA: Towards Difficulty-Stratified Time Series Reasoning from Visual Perception to Semantics},
  author={Yueyang Ding, HaoPeng Zhang, Rui Dai, Yi Wang, Tianyu Zong, Kaikui Liu, Xiangxiang Chu},
  journal={arxiv preprint arxiv: 2604.17295},
  year={2026}
}

Licensing

This dataset is released for research purposes.

Dataset Creator

Created as part of the LLaTiSA project.

Contact

For issues or questions regarding the dataset, please open an issue on the HuggingFace Hub.

Downloads last month
216

Paper for November-Rain/HiTSR