Hugging Face's logo

Datasets:
blue-orbit-472
/
IteractBench

License:
Dataset card Data Studio Files
xet
 Community
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 2 new columns ({'Seed', 'TeamID'}) and 7 missing columns ({'WLoc', 'DayNum', 'NumOT', 'WTeamID', 'WScore', 'LScore', 'LTeamID'}).

This happened while the csv dataset builder was generating data using

hf://datasets/blue-orbit-472/IteractBench/junesdata-loo-validation-lr-baseline-harbor-v2/environment/data/march-machine-learning-mania-2026/MNCAATourneySeeds.csv (at revision dc85120ed0e9f97819391f3b9d3e5f0d9478ccb8), ['hf://datasets/blue-orbit-472/IteractBench@dc85120ed0e9f97819391f3b9d3e5f0d9478ccb8/junesdata-loo-validation-lr-baseline-harbor-v2/environment/data/march-machine-learning-mania-2026/MNCAATourneyCompactResults.csv', 'hf://datasets/blue-orbit-472/IteractBench@dc85120ed0e9f97819391f3b9d3e5f0d9478ccb8/junesdata-loo-validation-lr-baseline-harbor-v2/environment/data/march-machine-learning-mania-2026/MNCAATourneySeeds.csv', 'hf://datasets/blue-orbit-472/IteractBench@dc85120ed0e9f97819391f3b9d3e5f0d9478ccb8/junesdata-loo-validation-lr-baseline-harbor-v2/environment/data/march-machine-learning-mania-2026/MRegularSeasonCompactResults.csv', 'hf://datasets/blue-orbit-472/IteractBench@dc85120ed0e9f97819391f3b9d3e5f0d9478ccb8/junesdata-loo-validation-lr-baseline-harbor-v2/environment/data/march-machine-learning-mania-2026/WNCAATourneyCompactResults.csv', 'hf://datasets/blue-orbit-472/IteractBench@dc85120ed0e9f97819391f3b9d3e5f0d9478ccb8/junesdata-loo-validation-lr-baseline-harbor-v2/environment/data/march-machine-learning-mania-2026/WNCAATourneySeeds.csv', 'hf://datasets/blue-orbit-472/IteractBench@dc85120ed0e9f97819391f3b9d3e5f0d9478ccb8/junesdata-loo-validation-lr-baseline-harbor-v2/environment/data/march-machine-learning-mania-2026/WRegularSeasonCompactResults.csv']

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/python3.14/site-packages/datasets/builder.py", line 1848, in _prepare_split_single
                  writer.write_table(table)
                  ~~~~~~~~~~~~~~~~~~^^^^^^^
                File "/usr/local/lib/python3.14/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/python3.14/site-packages/datasets/arrow_writer.py", line 773, in _write_table
                  pa_table = table_cast(pa_table, self._schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              Season: int64
              Seed: string
              TeamID: int64
              -- schema metadata --
              pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 606
              to
              {'Season': Value('int64'), 'DayNum': Value('int64'), 'WTeamID': Value('int64'), 'WScore': Value('int64'), 'LTeamID': Value('int64'), 'LScore': Value('int64'), 'WLoc': Value('string'), 'NumOT': Value('int64')}
              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 1369, in compute_config_parquet_and_info_response
                  parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
                                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      builder, max_dataset_size_bytes=max_dataset_size_bytes
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
                  builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
                  ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~^
                      gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  ):
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1850, in _prepare_split_single
                  raise DatasetGenerationCastError.from_cast_error(
                  ...<4 lines>...
                  )
              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 2 new columns ({'Seed', 'TeamID'}) and 7 missing columns ({'WLoc', 'DayNum', 'NumOT', 'WTeamID', 'WScore', 'LScore', 'LTeamID'}).
              
              This happened while the csv dataset builder was generating data using
              
              hf://datasets/blue-orbit-472/IteractBench/junesdata-loo-validation-lr-baseline-harbor-v2/environment/data/march-machine-learning-mania-2026/MNCAATourneySeeds.csv (at revision dc85120ed0e9f97819391f3b9d3e5f0d9478ccb8), ['hf://datasets/blue-orbit-472/IteractBench@dc85120ed0e9f97819391f3b9d3e5f0d9478ccb8/junesdata-loo-validation-lr-baseline-harbor-v2/environment/data/march-machine-learning-mania-2026/MNCAATourneyCompactResults.csv', 'hf://datasets/blue-orbit-472/IteractBench@dc85120ed0e9f97819391f3b9d3e5f0d9478ccb8/junesdata-loo-validation-lr-baseline-harbor-v2/environment/data/march-machine-learning-mania-2026/MNCAATourneySeeds.csv', 'hf://datasets/blue-orbit-472/IteractBench@dc85120ed0e9f97819391f3b9d3e5f0d9478ccb8/junesdata-loo-validation-lr-baseline-harbor-v2/environment/data/march-machine-learning-mania-2026/MRegularSeasonCompactResults.csv', 'hf://datasets/blue-orbit-472/IteractBench@dc85120ed0e9f97819391f3b9d3e5f0d9478ccb8/junesdata-loo-validation-lr-baseline-harbor-v2/environment/data/march-machine-learning-mania-2026/WNCAATourneyCompactResults.csv', 'hf://datasets/blue-orbit-472/IteractBench@dc85120ed0e9f97819391f3b9d3e5f0d9478ccb8/junesdata-loo-validation-lr-baseline-harbor-v2/environment/data/march-machine-learning-mania-2026/WNCAATourneySeeds.csv', 'hf://datasets/blue-orbit-472/IteractBench@dc85120ed0e9f97819391f3b9d3e5f0d9478ccb8/junesdata-loo-validation-lr-baseline-harbor-v2/environment/data/march-machine-learning-mania-2026/WRegularSeasonCompactResults.csv']
              
              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.

Season
int64
DayNum
int64
WTeamID
int64
WScore
int64
LTeamID
int64
LScore
int64
WLoc
string
NumOT
int64
1,985
136
1,116
63
1,234
54
N
0
1,985
136
1,120
59
1,345
58
N
0
1,985
136
1,207
68
1,250
43
N
0
1,985
136
1,229
58
1,425
55
N
0
1,985
136
1,242
49
1,325
38
N
0
1,985
136
1,246
66
1,449
58
N
0
1,985
136
1,256
78
1,338
54
N
0
1,985
136
1,260
59
1,233
58
N
0
1,985
136
1,314
76
1,292
57
N
0
1,985
136
1,323
79
1,333
70
N
0
1,985
136
1,326
75
1,235
64
N
0
1,985
136
1,328
96
1,299
83
N
0
1,985
136
1,374
85
1,330
68
N
0
1,985
136
1,385
83
1,380
59
N
0
1,985
136
1,396
60
1,439
57
N
0
1,985
136
1,424
85
1,361
80
N
0
1,985
137
1,104
50
1,112
41
N
0
1,985
137
1,130
55
1,403
53
N
0
1,985
137
1,181
75
1,337
62
N
0
1,985
137
1,208
67
1,455
59
N
0
1,985
137
1,210
65
1,273
58
N
0
1,985
137
1,228
76
1,318
57
N
0
1,985
137
1,268
69
1,275
68
N
1
1,985
137
1,272
67
1,335
55
N
0
1,985
137
1,276
59
1,192
55
N
0
1,985
137
1,298
78
1,261
55
N
0
1,985
137
1,301
65
1,305
56
N
0
1,985
137
1,393
70
1,177
65
N
0
1,985
137
1,412
70
1,277
68
N
0
1,985
137
1,431
79
1,409
75
N
0
1,985
137
1,433
81
1,267
65
N
0
1,985
137
1,437
51
1,173
49
N
0
1,985
138
1,120
66
1,242
64
N
0
1,985
138
1,207
63
1,396
46
N
0
1,985
138
1,246
64
1,424
61
N
0
1,985
138
1,256
79
1,326
67
N
0
1,985
138
1,260
70
1,374
57
N
0
1,985
138
1,314
60
1,323
58
N
0
1,985
138
1,328
75
1,229
69
N
0
1,985
138
1,385
68
1,116
65
N
0
1,985
139
1,104
63
1,433
59
N
0
1,985
139
1,130
74
1,181
73
N
0
1,985
139
1,210
70
1,393
53
N
0
1,985
139
1,228
74
1,208
58
N
0
1,985
139
1,268
64
1,298
59
N
0
1,985
139
1,272
67
1,412
66
N
1
1,985
139
1,301
86
1,431
73
N
0
1,985
139
1,437
59
1,276
55
N
0
1,985
143
1,207
65
1,260
53
N
0
1,985
143
1,210
61
1,228
53
N
0
1,985
143
1,272
59
1,130
57
N
0
1,985
143
1,328
86
1,256
84
N
1
1,985
144
1,301
61
1,104
55
N
0
1,985
144
1,314
62
1,120
56
N
0
1,985
144
1,385
86
1,246
70
N
0
1,985
144
1,437
46
1,268
43
N
0
1,985
145
1,207
60
1,210
54
N
0
1,985
145
1,272
63
1,328
61
N
0
1,985
146
1,385
69
1,301
60
N
0
1,985
146
1,437
56
1,314
44
N
0
1,985
152
1,207
77
1,385
59
N
0
1,985
152
1,437
52
1,272
45
N
0
1,985
154
1,437
66
1,207
64
N
0
1,986
136
1,133
83
1,431
65
N
0
1,986
136
1,177
72
1,438
68
N
0
1,986
136
1,181
85
1,290
78
N
0
1,986
136
1,207
70
1,403
64
N
0
1,986
136
1,210
68
1,265
53
N
0
1,986
136
1,242
71
1,299
46
N
0
1,986
136
1,257
93
1,180
73
N
0
1,986
136
1,261
94
1,345
87
N
1
1,986
136
1,272
95
1,123
63
N
0
1,986
136
1,277
72
1,449
70
N
0
1,986
136
1,314
84
1,428
72
N
0
1,986
136
1,328
80
1,318
74
N
0
1,986
136
1,330
72
1,452
64
N
0
1,986
136
1,396
61
1,239
50
N
1
1,986
136
1,412
66
1,281
64
N
0
1,986
136
1,437
71
1,439
62
N
0
1,986
137
1,104
97
1,462
80
N
0
1,986
137
1,114
90
1,323
83
N
0
1,986
137
1,120
73
1,112
63
N
0
1,986
137
1,156
83
1,231
79
N
0
1,986
137
1,228
75
1,193
51
N
0
1,986
137
1,235
81
1,275
79
N
1
1,986
137
1,246
75
1,172
55
N
0
1,986
137
1,268
69
1,337
64
N
0
1,986
137
1,276
70
1,103
64
N
0
1,986
137
1,298
87
1,409
68
N
0
1,986
137
1,301
66
1,234
64
N
0
1,986
137
1,385
83
1,286
74
N
0
1,986
137
1,386
60
1,350
59
N
0
1,986
137
1,393
101
1,135
52
N
0
1,986
137
1,424
74
1,419
51
N
0
1,986
137
1,443
67
1,304
59
N
0
1,986
138
1,177
74
1,328
69
N
0
1,986
138
1,181
89
1,330
61
N
0
1,986
138
1,210
66
1,437
61
N
0
1,986
138
1,242
65
1,396
43
N
0
1,986
138
1,257
82
1,133
68
N
0
End of preview.

Iteract-Bench

Benchmark data for ITERACT-BENCH: Benchmarking Agents for Interactive Data Analysis Workflows.

The code for running the benchmark is at https://anonymous.4open.science/r/Iteract-Bench-CDFD.

Each directory is one interactive data-analysis task. A user simulator discloses analytical requirements during the episode. An independent verifier later checks stage answers against gold outputs that the simulator never sees.

Statistics

Statistic Value
Tasks 86
Analytical stages 554
Data files 727
Stage types 12
Stages per task (min / avg / max) 2 / 6.4 / 16
DAG depth (min / avg / max) 2 / 3.2 / 8
Stages with upstream dependencies 80.9%
Non-chain DAGs 83.7%

Domain share of the 86 tasks:

  • Sports & competitions 24.4%
  • Education & labor 15.1%
  • Other 14.0%
  • Energy & environment 10.5%
  • Customer & churn 9.3%
  • Finance & markets 8.1%
  • Industry & transport 7.0%
  • Health & fitness 5.8%
  • Media & entertainment 5.8%

Task layout

Each task is a Harbor task: instruction.md, task.toml, environment/, and tests/.

A task is T = (D, G, τ, s0, U). Each stage is ni = (Di, Ii, Fi, Yi, Vi) (Section 2.1).

Path Contents Paper
environment/data/ Shared CSV tables Data environment D
tests/turn_spec.json DAG edges and order; per-stage surface, hidden_requirements, and answer_schema. No gold values G, τ; Initial Request Ii; Full Task Specification Fi; output space Yi
tests/user_scenario.md Simulator policy: ASSIGN, CLARIFY, CORRECT, COMPLETE User simulator U
instruction.md Agent protocol (ask_question, submit_response) Not the analytical specification
harbor_ir.json Authored IR used to compile the package: goal, turns, data manifest, hidden requirements Source of Ii and Fi
tests/ground_truth.json Gold stage outputs from reference execution. Verifier-only; not mounted for the agent Gold y*
tests/evaluate.py, tests/exact_match_eval.py, tests/dag_evaluate.py Post-episode exact match of structured answers Stage verifier Vi
task.toml, environment/Dockerfile*, environment/runtime-server/task_config.json Harbor sandbox and MCP runtime Execution environment

In on-demand mode the simulator's first ASSIGN shows only the surface (Ii). Hidden requirements in Fi are disclosed when the agent asks or a submission conflicts with the specification. Stage completion is not correctness; Vi runs after the episode.
Downloads last month
375
Total file size:
3.1 GB