Dataset Viewer
Auto-converted to Parquet Duplicate
slug
stringlengths
7
41
wave
stringclasses
1 value
url
stringlengths
43
77
task_type
stringclasses
2 values
n_features
int64
4
4.99k
n_train
int64
200
720k
n_test
int64
50
1.21M
true_target_column_in_ambig
stringclasses
2 values
cv_true
float64
0
1
cv_decoy
float64
-0
1
cv_ratio_decoy_over_true
float64
-2.14
1.9
abs_correlation_truth_vs_decoy
float64
0
0.46
bike-sharing-demand
dsbench_original
https://www.kaggle.com/competitions/bike-sharing-demand
regression
8
8,708
2,178
val_1
0.337837
0.642299
1.901209
0.008289
cat-in-the-dat
dsbench_original
https://www.kaggle.com/competitions/cat-in-the-dat
classification
23
240,000
60,000
val_1
0.596016
0.496015
0.832219
0.003801
cat-in-the-dat-ii
dsbench_original
https://www.kaggle.com/competitions/cat-in-the-dat-ii
classification
23
480,000
120,000
val_2
0.618799
0.504609
0.815465
0.002474
dont-overfit-ii
dsbench_original
https://www.kaggle.com/competitions/dont-overfit-ii
classification
300
200
50
val_1
0.540399
0.560665
1.037501
0.018525
instant-gratification
dsbench_original
https://www.kaggle.com/competitions/instant-gratification
classification
256
209,715
52,429
val_2
0.503094
0.552716
1.098634
0.003839
playground-series-s3e1
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e1
regression
8
29,709
7,428
val_1
0.745288
0.743674
0.997834
0.037061
playground-series-s3e10
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e10
classification
8
94,051
23,513
val_1
0.994013
0.999716
1.005738
0.098588
playground-series-s3e11
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e11
regression
15
288,268
72,068
val_1
0.063637
-0.001539
-0.024188
0.000374
playground-series-s3e12
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e12
classification
6
331
83
val_1
0.759015
0.720103
0.948733
0.14217
playground-series-s3e13
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e13
classification
64
565
142
val_2
0.284992
0.281427
0.987491
0.055471
playground-series-s3e14
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e14
regression
16
12,231
3,058
val_2
0.818535
0.824783
1.007634
0.009807
playground-series-s3e16
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e16
regression
8
59,240
14,811
val_2
0.565977
0.622398
1.099687
0.45524
playground-series-s3e17
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e17
classification
12
109,143
27,286
val_2
0.949359
0.967128
1.018717
0.032205
playground-series-s3e2
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e2
classification
10
12,243
3,061
val_2
0.874828
0.847805
0.969111
0.090185
playground-series-s3e20
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e20
regression
75
63,218
15,805
val_1
0.768857
0.674041
0.876679
0.005505
playground-series-s3e22
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e22
classification
27
988
247
val_1
0.652817
0.633579
0.970531
0.143749
playground-series-s3e23
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e23
classification
21
81,410
20,353
val_1
0.790692
0.780042
0.98653
0.178169
playground-series-s3e24
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e24
classification
22
127,404
31,852
val_1
0.852413
0.836573
0.981418
0.004604
playground-series-s3e25
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e25
regression
11
8,325
2,082
val_1
0.46594
0.568254
1.219586
0.101616
playground-series-s3e3
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e3
classification
33
1,341
336
val_2
0.782554
0.74359
0.950208
0.015579
playground-series-s3e4
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e4
classification
30
175,303
43,826
val_2
0.572253
0.491073
0.858141
0.003404
playground-series-s3e5
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e5
classification
11
1,644
412
val_1
0.569951
0.587591
1.03095
0.018166
playground-series-s3e6
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e6
regression
16
18,184
4,546
val_2
0.996033
0.977547
0.98144
0.002171
playground-series-s3e7
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e7
classification
17
33,680
8,420
val_2
0.887472
0.900579
1.014769
0.027498
playground-series-s3e8
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e8
regression
9
154,858
38,715
val_1
0.91694
0.898852
0.980273
0.159464
playground-series-s3e9
dsbench_original
https://www.kaggle.com/competitions/playground-series-s3e9
regression
8
4,325
1,082
val_1
0.419683
0.576896
1.374599
0.063882
playground-series-s4e1
dsbench_original
https://www.kaggle.com/competitions/playground-series-s4e1
classification
12
132,027
33,007
val_1
0.872134
0.871843
0.999666
0.005479
playground-series-s4e2
dsbench_original
https://www.kaggle.com/competitions/playground-series-s4e2
classification
16
16,606
4,152
val_2
0.892027
0.738769
0.828191
0.207493
playground-series-s4e4
dsbench_original
https://www.kaggle.com/competitions/playground-series-s4e4
classification
8
72,492
18,123
val_2
0.33405
0.3426
1.025596
0.223497
playground-series-s4e5
dsbench_original
https://www.kaggle.com/competitions/playground-series-s4e5
regression
20
500,000
223,592
val_2
0.67
0.671
1.001493
0.145
playground-series-s4e6
dsbench_original
https://www.kaggle.com/competitions/playground-series-s4e6
classification
36
61,214
15,304
val_1
0.82505
0.8321
1.008545
0.003096
porto-seguro-safe-driver-prediction
dsbench_original
https://www.kaggle.com/competitions/porto-seguro-safe-driver-prediction
classification
57
476,169
119,043
val_2
0.592538
0.577459
0.974552
0.000468
santander-customer-satisfaction
dsbench_original
https://www.kaggle.com/competitions/santander-customer-satisfaction
classification
369
60,816
15,204
val_1
0.820757
0.823001
1.002734
0.00486
santander-customer-transaction-prediction
dsbench_original
https://www.kaggle.com/competitions/santander-customer-transaction-prediction
classification
201
160,000
40,000
val_1
0.801483
0.813679
1.015217
0.001583
santander-value-prediction-challenge
dsbench_original
https://www.kaggle.com/competitions/santander-value-prediction-challenge
regression
4,991
3,567
892
val_1
0.34
0.55
1.617647
0.161
spaceship-titanic
dsbench_original
https://www.kaggle.com/competitions/spaceship-titanic
classification
12
6,954
1,739
val_1
0.849831
0.848204
0.998085
0.039447
tabular-playground-series-apr-2021
dsbench_original
https://www.kaggle.com/competitions/tabular-playground-series-apr-2021
classification
10
80,000
20,000
val_2
0.73487
0.748963
1.019179
0.007829
tabular-playground-series-aug-2021
dsbench_original
https://www.kaggle.com/competitions/tabular-playground-series-aug-2021
regression
100
200,000
50,000
val_1
0.00216
-0.004627
-2.142484
0.002315
tabular-playground-series-aug-2022
dsbench_original
https://www.kaggle.com/competitions/tabular-playground-series-aug-2022
classification
24
21,256
5,314
val_1
0.569884
0.588518
1.032699
0.004173
tabular-playground-series-feb-2021
dsbench_original
https://www.kaggle.com/competitions/tabular-playground-series-feb-2021
regression
24
240,000
60,000
val_1
0.036759
-0.002176
-0.059191
0.002149
tabular-playground-series-feb-2022
dsbench_original
https://www.kaggle.com/competitions/tabular-playground-series-feb-2022
classification
286
66,900
16,725
val_2
0.8825
0.83605
0.947365
0.027978
tabular-playground-series-jan-2021
dsbench_original
https://www.kaggle.com/competitions/tabular-playground-series-jan-2021
regression
14
240,000
60,000
val_1
0.040736
-0.003862
-0.094793
0.000183
tabular-playground-series-mar-2021
dsbench_original
https://www.kaggle.com/competitions/tabular-playground-series-mar-2021
classification
30
240,000
60,000
val_1
0.785725
0.795839
1.012871
0.055554
tabular-playground-series-mar-2022
dsbench_original
https://www.kaggle.com/competitions/tabular-playground-series-mar-2022
regression
4
500,000
169,767
val_1
0.229
0.269
1.174672
0.028472
tabular-playground-series-may-2022
dsbench_original
https://www.kaggle.com/competitions/tabular-playground-series-may-2022
classification
31
720,000
180,000
val_1
0.842237
0.857426
1.018034
0.006274
tabular-playground-series-nov-2021
dsbench_original
https://www.kaggle.com/competitions/tabular-playground-series-nov-2021
classification
100
480,000
120,000
val_2
0.70068
0.736044
1.050471
0.000329
tabular-playground-series-oct-2021
dsbench_original
https://www.kaggle.com/competitions/tabular-playground-series-oct-2021
classification
285
500,000
200,000
val_2
0.835491
0.832067
0.995901
0.005712
tabular-playground-series-sep-2021
dsbench_original
https://www.kaggle.com/competitions/tabular-playground-series-sep-2021
classification
118
416,946
104,237
val_2
0.773953
0.780326
1.008235
0.000176
titanic
dsbench_original
https://www.kaggle.com/competitions/titanic
classification
10
712
179
val_1
0.734204
0.733887
0.999569
0.01869
tmdb-box-office-prediction
dsbench_original
https://www.kaggle.com/competitions/tmdb-box-office-prediction
regression
21
2,400
600
val_1
0.592278
0.632424
1.067783
0.321406
ventilator-pressure-prediction
dsbench_original
https://www.kaggle.com/competitions/ventilator-pressure-prediction
regression
6
500,000
1,207,200
val_1
0.726369
0.744277
1.024654
0.043289

Ambig-DS-T: Target Ambiguity Benchmark

A benchmark for measuring how well data-science agents handle ambiguous prediction targets in tabular Kaggle competitions.

Each task is a Kaggle competition derived from DSBench. For every task we provide two prompt variants — one in which the target column is named, and one in which the target is hidden behind two candidate columns. The agent must select and predict the true target; submissions are graded by the original competition metric using DSBench's per-task evaluator.

The benchmark contains 51 paired tasks (33 classification, 18 regression).

Variants

Variant File Description
Full tasks/{slug}/task.txt Original task description — names the target column verbatim and lists features under their semantic names.
Ambiguous tasks/{slug}/task_ambig.txt Same description with target identity hidden: feature names are anonymized to f_01, f_02, …, the original target name is removed, and training data exposes two candidate target columns val_1 and val_2. Exactly one is the true target; the other is a feature-predictable decoy with the same marginal distribution.

The Full arm establishes the upper baseline; the Ambiguous arm is the diagnostic condition. The failure mode the benchmark measures: an agent that silently picks the decoy column will train a model with normal-looking cross-validation behaviour but score poorly on the held-out test, because the decoy is constructed to be approximately uncorrelated with the true target.

Layout

tasks/
  {slug}/
    task.txt              # Full task description (target named)
    task_ambig.txt        # Target-redacted version (val_1 + val_2 decoy)
    eval.py               # Per-task evaluator (DSBench CLI; metric varies by task)
    _manifest.json        # Provenance + decoy recipe + diagnostics
tasks.csv                 # Flat 51-row index (slug, task_type, n_train, n_test, oracle target column, …)

Setup: getting the competition data

This dataset contains prompts, evaluators, and decoy recipes only — not the Kaggle competition data (train/test CSVs). To respect each competition's terms of use, the redistributed contents are deterministic recipes over data the user must download themselves.

To run the benchmark:

  1. Accept each competition's rules in your browser (the source.rules_url field of _manifest.json links straight there).
  2. Download the data via the official Kaggle CLI (kaggle competitions download -c <slug>).
  3. Use the build script (published separately on GitHub) to apply the deterministic recipe in _manifest.json.ambig_recipe. This rebuilds the ambig CSVs bit-identically using the recorded seeds.master and seeds.decoy.

Manifest

tasks/{slug}/_manifest.json records, per task:

Section Purpose
source platform, url, rules_url, wave (all 51 are dsbench_original).
task task_type (classification/regression), id_column, true_target_column_in_ambig (oracle: val_1 or val_2), decoy_column_in_ambig, original_target_name, n_features, n_train, n_test.
ambig_recipe Deterministic decoy generation. method (e.g. rank_map_lowcorr_pool+label_noise), full feature_map (original→anon), anon_feature_columns, decoy_pool_anon_features and their decoy_pool_abs_spearman_with_truth, label-noise fractions, seeds.master / seeds.decoy.
diagnostics cv_true, cv_decoy, cv_ratio_decoy_over_true, correlation_truth_vs_decoy, marginal_match_exact. The decoy is calibrated so that its values are approximately uncorrelated with the true target while remaining roughly as feature-predictable.
eval Pointer to the local eval.py and its CLI signature.

The task.true_target_column_in_ambig field is the clarification oracle's source of truth: in the clarify experimental condition, an answerer LLM resolves the agent's questions about which column to predict using only this field. It is intentionally never given to the agent in the ambig (no-clarify) condition.

Example task and diagnostics (from bike-sharing-demand):

{
  "task": {
    "task_type": "regression",
    "id_column": "datetime",
    "true_target_column_in_ambig": "val_1",
    "decoy_column_in_ambig": "val_2",
    "original_target_name": "count",
    "n_features": 8,
    "n_train": 8708,
    "n_test": 2178
  },
  "diagnostics": {
    "cv_true": 0.34,
    "cv_decoy": 0.64,
    "cv_ratio_decoy_over_true": 1.90,
    "correlation_truth_vs_decoy": 0.01,
    "marginal_match_exact": true
  }
}

cv_decoy matches cv_true in feature-predictability (median ratio across the benchmark $\approx 1.0$) while correlation_truth_vs_decoy ≈ 0 (median $|\rho_{\mathrm{Spearman}}| = 0.017$): the two columns are equally learnable but almost orthogonal, so an agent that picks the wrong column still gets a high CV score on data that is unrelated to the true target.

Evaluating a submission

Every eval.py accepts the same DSBench-style CLI:

python eval.py --answer_file data/test_answer.csv \
               --predict_file my_submission.csv \
               --path out --name <slug>

…and writes a single float (the competition's original metric — RMSLE / AUC / RMSE / accuracy / …) to out/<slug>/result.txt.

Tasks (51)

# Competition Type True column n_train n_test n_features
1 bike-sharing-demand regression val_1 8,708 2,178 8
2 cat-in-the-dat classification val_1 240,000 60,000 23
3 cat-in-the-dat-ii classification val_2 480,000 120,000 23
4 dont-overfit-ii classification val_1 200 50 300
5 instant-gratification classification val_2 209,715 52,429 256
6 playground-series-s3e1 regression val_1 29,709 7,428 8
7 playground-series-s3e10 classification val_1 94,051 23,513 8
8 playground-series-s3e11 regression val_1 288,268 72,068 15
9 playground-series-s3e12 classification val_1 331 83 6
10 playground-series-s3e13 classification val_2 565 142 64
11 playground-series-s3e14 regression val_2 12,231 3,058 16
12 playground-series-s3e16 regression val_2 59,240 14,811 8
13 playground-series-s3e17 classification val_2 109,143 27,286 12
14 playground-series-s3e2 classification val_2 12,243 3,061 10
15 playground-series-s3e20 regression val_1 63,218 15,805 75
16 playground-series-s3e22 classification val_1 988 247 27
17 playground-series-s3e23 classification val_1 81,410 20,353 21
18 playground-series-s3e24 classification val_1 127,404 31,852 22
19 playground-series-s3e25 regression val_1 8,325 2,082 11
20 playground-series-s3e3 classification val_2 1,341 336 33
21 playground-series-s3e4 classification val_2 175,303 43,826 30
22 playground-series-s3e5 classification val_1 1,644 412 11
23 playground-series-s3e6 regression val_2 18,184 4,546 16
24 playground-series-s3e7 classification val_2 33,680 8,420 17
25 playground-series-s3e8 regression val_1 154,858 38,715 9
26 playground-series-s3e9 regression val_1 4,325 1,082 8
27 playground-series-s4e1 classification val_1 132,027 33,007 12
28 playground-series-s4e2 classification val_2 16,606 4,152 16
29 playground-series-s4e4 classification val_2 72,492 18,123 8
30 playground-series-s4e5 regression val_2 500,000 223,592 20
31 playground-series-s4e6 classification val_1 61,214 15,304 36
32 porto-seguro-safe-driver-prediction classification val_2 476,169 119,043 57
33 santander-customer-satisfaction classification val_1 60,816 15,204 369
34 santander-customer-transaction-prediction classification val_1 160,000 40,000 201
35 santander-value-prediction-challenge regression val_1 3,567 892 4,991
36 spaceship-titanic classification val_1 6,954 1,739 12
37 tabular-playground-series-apr-2021 classification val_2 80,000 20,000 10
38 tabular-playground-series-aug-2021 regression val_1 200,000 50,000 100
39 tabular-playground-series-aug-2022 classification val_1 21,256 5,314 24
40 tabular-playground-series-feb-2021 regression val_1 240,000 60,000 24
41 tabular-playground-series-feb-2022 classification val_2 66,900 16,725 286
42 tabular-playground-series-jan-2021 regression val_1 240,000 60,000 14
43 tabular-playground-series-mar-2021 classification val_1 240,000 60,000 30
44 tabular-playground-series-mar-2022 regression val_1 500,000 169,767 4
45 tabular-playground-series-may-2022 classification val_1 720,000 180,000 31
46 tabular-playground-series-nov-2021 classification val_2 480,000 120,000 100
47 tabular-playground-series-oct-2021 classification val_2 500,000 200,000 285
48 tabular-playground-series-sep-2021 classification val_2 416,946 104,237 118
49 titanic classification val_1 712 179 10
50 tmdb-box-office-prediction regression val_1 2,400 600 21
51 ventilator-pressure-prediction regression val_1 500,000 1,207,200 6

The True column column is the answer the clarification oracle returns when an agent asks which of val_1/val_2 to predict; it is never shown to the agent in the ambig condition.

Citation

@article{ambig-ds-2026,
  title  = {Ambig-DS: Diagnosing Unflagged Misframings in Data-Science Agents},
  year   = {2026},
  note   = {NeurIPS 2026 Datasets \& Benchmarks submission (under review)}
}

License

The contents of this repository (prompts, manifests, task index, decoy recipes) are released under CC-BY-NC-4.0, inheriting the non-commercial research-use restriction from the upstream DSBench dataset terms (DSBench code is MIT). The task.txt files are factual paraphrases of publicly available Kaggle competition descriptions; the task_ambig.txt files, the decoy-generation recipes in _manifest.json.ambig_recipe, and the per-task diagnostics are original contributions.

The per-task eval.py evaluators are redistributed unchanged from DSBench (Jing et al., 2024) so that grading remains bit-identical to upstream. Some still contain inline comments in Chinese — these are upstream artefacts.

The underlying Kaggle competition datasets are not redistributed here. They must be downloaded separately via the Kaggle API and remain subject to each competition's individual rules and terms of use.

Downloads last month
73