Datasets:

EdwardoSunny
/

mle-rubrics-filtered-final

rubric_id stringlengths 12 12	text stringlengths 794 1.74k	retrieval_text stringlengths 107 309	input_condition stringclasses 2 values	family stringclasses 1 value	source_instance_id stringlengths 25 68	source_repo stringclasses 3 values	source_dataset stringclasses 3 values	library stringclasses 1 value	manifest_rank int64 0 2.75k	title stringlengths 37 111	prompt_block stringlengths 794 1.74k	weight int64 0 434	error_weight int64 0 303	outcome_weight int64 0 246	n_rollouts int64 1 1.42k	n_merged int64 1 1.42k	evidence_ids listlengths 1 50
7976cfec37dd	### Inline heredoc used for analysis despite an explicit .py-file requirement - Kind: operational - Applies when: `code` -- instructions state code must live in `.py` files executed with `python file.py` - Pattern: verification or analysis logic is piped into the interpreter via `python << 'EOF' ... EOF`, s...	Inline heredoc used for analysis despite an explicit .py-file requirement -- `code` -- instructions state code must live in `.py` files executed with `python file.py`	code	mle_clean_v3	da-pool-clean/dsbench-bike-sharing-demand@s2#5	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	0	Inline heredoc used for analysis despite an explicit .py-file requirement	### Inline heredoc used for analysis despite an explicit .py-file requirement - Kind: operational - Applies when: `code` -- instructions state code must live in `.py` files executed with `python file.py` - Pattern: verification or analysis logic is piped into the interpreter via `python << 'EOF' ... EOF`, s...	434	301	133	1,419	1,420	[ "da-pool-clean/dsbench-bike-sharing-demand@s2#5", "da-pool-clean/dsbench-nlp-getting-started#0", "da-pool-clean/dsbench-tmdb-box-office-prediction#1", "da-pool-clean/dsbench-nlp-getting-started@s3#4", "da-pool-clean/dsbench-santander-value-prediction-challenge@s2#3", "da-pool-clean/dsbench-dont-overfit-ii...
ce777eafc02e	### Gradient-boosting library imported without checking the installed package list - Kind: operational - Applies when: `code` -- the script imports a third-party modelling library (xgboost, lightgbm, catboost, torch, prophet) that the task description or environment listing never mentions. - Pattern: the im...	Gradient-boosting library imported without checking the installed package list -- `code` -- the script imports a third-party modelling library (xgboost, lightgbm, catboost, torch, prophet) that the task description or environment listing never mentions.	code	mle_clean_v3	da-pool-clean/dsbench-demand-forecasting-kernels-only@s2#0	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	1	Gradient-boosting library imported without checking the installed package list	### Gradient-boosting library imported without checking the installed package list - Kind: operational - Applies when: `code` -- the script imports a third-party modelling library (xgboost, lightgbm, catboost, torch, prophet) that the task description or environment listing never mentions. - Pattern: the im...	303	303	0	229	231	[ "da-pool-clean/dsbench-demand-forecasting-kernels-only@s2#0", "da-pool-clean/dsbench-commonlitreadabilityprize#0", "da-pool-clean/dsbench-demand-forecasting-kernels-only#0", "da-pool-clean/dsbench-tmdb-box-office-prediction@s2#0", "da-pool-clean/dsbench-santander-customer-transaction-prediction#0", "da-po...
3c973d860095	### Submitting prose instead of the required artifact content - Kind: methodology - Applies when: `task` -- the instructions require the answer written in an exact requested format to a specific answer file while the graded artifact is a CSV. - Pattern: The answer file is filled with a narrative summary ("a...	Submitting prose instead of the required artifact content -- `task` -- the instructions require the answer written in an exact requested format to a specific answer file while the graded artifact is a CSV.	task	mle_clean_v3	da-pool-clean-da/dacode-dm-csv-011#4	da-code	clean:methodology	mle-rubrics-clean-v3-oraclepool	2	Submitting prose instead of the required artifact content	### Submitting prose instead of the required artifact content - Kind: methodology - Applies when: `task` -- the instructions require the answer written in an exact requested format to a specific answer file while the graded artifact is a CSV. - Pattern: The answer file is filled with a narrative summary ("a...	249	3	246	238	255	[ "da-pool-clean-da/dacode-dm-csv-011#4", "da-pool-clean-da/dacode-ml-cluster-016#5", "da-pool-clean-da/dacode-ml-cluster-014#4", "da-pool-clean-da/dacode-plot-line-006#3", "da-pool-clean-da/dacode-plot-line-006#4", "da-pool-clean-da/dacode-plot-bar-005#4", "da-pool-clean-da/dacode-di-text-003#1", "da-p...
b49c6bda6bc0	### Single-threaded sklearn boosting sized without a runtime estimate - Kind: operational - Applies when: `code` -- `GradientBoostingClassifier` (not the Hist/`n_jobs` variants) is fit on a wide table inside one command with a hard wall-clock limit - Pattern: `GradientBoostingClassifier(n_estimators=N, max_...	Single-threaded sklearn boosting sized without a runtime estimate -- `code` -- `GradientBoostingClassifier` (not the Hist/`n_jobs` variants) is fit on a wide table inside one command with a hard wall-clock limit	code	mle_clean_v3	da-pool-clean/dsbench-santander-customer-satisfaction#3	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	3	Single-threaded sklearn boosting sized without a runtime estimate	### Single-threaded sklearn boosting sized without a runtime estimate - Kind: operational - Applies when: `code` -- `GradientBoostingClassifier` (not the Hist/`n_jobs` variants) is fit on a wide table inside one command with a hard wall-clock limit - Pattern: `GradientBoostingClassifier(n_estimators=N, max_...	144	144	0	98	99	[ "da-pool-clean/dsbench-santander-customer-satisfaction#3", "da-pool-clean/dsbench-microsoft-malware-prediction@s2#5", "da-pool-clean/dsbench-santander-customer-satisfaction@s3#4", "da-pool-clean/dsbench-demand-forecasting-kernels-only#5", "da-pool-clean/dsbench-liverpool-ion-switching#3", "da-pool-clean/d...
86ae00d73ce3	### Trusting delimited-file headers without checking for BOM / stray encoding artifacts - Kind: operational - Applies when: `code` -- a CSV/TSV is read with default settings and columns are then referenced by literal name - Pattern: `pd.read_csv(path)` on a file whose first byte sequence is a UTF-8 BOM (vis...	Trusting delimited-file headers without checking for BOM / stray encoding artifacts -- `code` -- a CSV/TSV is read with default settings and columns are then referenced by literal name	code	mle_clean_v3	da-pool-clean-da/dabench-474#1	infiagent-dabench	clean:operational	mle-rubrics-clean-v3-oraclepool	4	Trusting delimited-file headers without checking for BOM / stray encoding artifacts	### Trusting delimited-file headers without checking for BOM / stray encoding artifacts - Kind: operational - Applies when: `code` -- a CSV/TSV is read with default settings and columns are then referenced by literal name - Pattern: `pd.read_csv(path)` on a file whose first byte sequence is a UTF-8 BOM (vis...	124	124	0	150	153	[ "da-pool-clean-da/dabench-474#1", "da-pool-clean-da/dabench-730#0", "da-pool-clean-da/dabench-234#1", "da-pool-clean-da/dabench-449#3", "da-pool-clean-da/dabench-450#2", "da-pool-clean-da/dabench-587#0", "da-pool-clean-da/dabench-447#0", "da-pool-clean-da/dabench-446#0", "da-pool-clean-da/dacode-dm-...
ab7069bf3458	### Fabricating substitute input data when the real data is not found - Kind: causal - Applies when: `code` -- a script generates random/synthetic rows and writes them into the task's data directory, or otherwise invents a test index. - Pattern: A `create_*_data.py` that `np.random`-generates train/test fra...	Fabricating substitute input data when the real data is not found -- `code` -- a script generates random/synthetic rows and writes them into the task's data directory, or otherwise invents a test index.	code	mle_clean_v3	da-pool-clean/dsbench-porto-seguro-safe-driver-prediction@s2#4	dsbench	clean:causal	mle-rubrics-clean-v3-oraclepool	6	Fabricating substitute input data when the real data is not found	### Fabricating substitute input data when the real data is not found - Kind: causal - Applies when: `code` -- a script generates random/synthetic rows and writes them into the task's data directory, or otherwise invents a test index. - Pattern: A `create_*_data.py` that `np.random`-generates train/test fra...	61	13	48	65	65	[ "da-pool-clean/dsbench-porto-seguro-safe-driver-prediction@s2#4", "da-pool-clean/dsbench-santander-customer-satisfaction@s2#1", "da-pool-clean/dsbench-liverpool-ion-switching#5", "da-pool-clean-da/dacode-data-sa-028#0", "da-pool-clean-da/dacode-plot-line-015#3", "da-pool-clean-da/dacode-plot-bar-007#2", ...
96031f143505	### Chained `df[col].fillna(..., inplace=True)` under modern pandas - Kind: operational - Applies when: `code` -- missing values are filled by calling an in-place method on a column selected out of a DataFrame. - Pattern: `df['x'].fillna(v, inplace=True)` (or similar chained in-place mutation) inside a loop...	Chained `df[col].fillna(..., inplace=True)` under modern pandas -- `code` -- missing values are filled by calling an in-place method on a column selected out of a DataFrame.	code	mle_clean_v3	da-pool-clean/dsbench-demand-forecasting-kernels-only@s2#1	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	7	Chained `df[col].fillna(..., inplace=True)` under modern pandas	### Chained `df[col].fillna(..., inplace=True)` under modern pandas - Kind: operational - Applies when: `code` -- missing values are filled by calling an in-place method on a column selected out of a DataFrame. - Pattern: `df['x'].fillna(v, inplace=True)` (or similar chained in-place mutation) inside a loop...	60	49	11	127	127	[ "da-pool-clean/dsbench-demand-forecasting-kernels-only@s2#1", "da-pool-clean/dsbench-microsoft-malware-prediction@s4#1", "da-pool-clean/dsbench-commonlitreadabilityprize@s5#1", "da-pool-clean/dsbench-cat-in-the-dat-ii@s7#1", "da-pool-clean/dsbench-demand-forecasting-kernels-only@s8#3", "da-pool-clean/dsbe...
3e683196c1fa	### No held-out validation before trusting the model - Kind: methodology - Applies when: `code` -- a script fits models and writes the submission in a single pass with no evaluation split - Pattern: models are fit on 100% of the labelled rows, predictions are written out, and the only "checks" are descripti...	No held-out validation before trusting the model -- `code` -- a script fits models and writes the submission in a single pass with no evaluation split	code	mle_clean_v3	da-pool-clean/dsbench-bike-sharing-demand@s2#0	dsbench	clean:methodology	mle-rubrics-clean-v3-oraclepool	8	No held-out validation before trusting the model	### No held-out validation before trusting the model - Kind: methodology - Applies when: `code` -- a script fits models and writes the submission in a single pass with no evaluation split - Pattern: models are fit on 100% of the labelled rows, predictions are written out, and the only "checks" are descripti...	47	1	46	81	81	[ "da-pool-clean/dsbench-bike-sharing-demand@s2#0", "da-pool-clean/dsbench-bike-sharing-demand#0", "da-pool-clean/dsbench-covid19-global-forecasting-week-1@s2#3", "da-pool-clean/dsbench-covid19-global-forecasting-week-1@s3#0", "da-pool-clean/dsbench-bike-sharing-demand@s4#2", "da-pool-clean/dsbench-covid19-...
75b1273fc3a5	### "Verification" that re-executes the same code path - Kind: methodology - Applies when: `code` -- a separate verify/check script is added before submitting. - Pattern: The verification script copies the same merges, the same revenue formula and the same filter as the analysis script, then prints the same...	"Verification" that re-executes the same code path -- `code` -- a separate verify/check script is added before submitting.	code	mle_clean_v3	da-pool-clean-da/dacode-dm-csv-011#3	da-code	clean:methodology	mle-rubrics-clean-v3-oraclepool	9	"Verification" that re-executes the same code path	### "Verification" that re-executes the same code path - Kind: methodology - Applies when: `code` -- a separate verify/check script is added before submitting. - Pattern: The verification script copies the same merges, the same revenue formula and the same filter as the analysis script, then prints the same...	46	1	45	79	79	[ "da-pool-clean-da/dacode-dm-csv-011#3", "da-pool-clean-da/dacode-data-sa-001#3", "da-pool-clean-da/dacode-ml-competition-005#1", "da-pool-clean-da/dacode-dm-csv-050#1", "da-pool-clean-da/dacode-plot-bar-005#3", "da-pool-clean-da/dacode-data-sa-028#4", "da-pool-clean-da/dabench-722#1", "da-pool-clean-d...
630488c85267	### Fitting a label/ordinal encoder on train only, then transforming test - Kind: operational - Applies when: `code` -- categorical columns are encoded with `LabelEncoder`/mapping fitted on the training frame and applied to the test frame - Pattern: `le.fit_transform(train[col]); le.transform(test[col])` fo...	Fitting a label/ordinal encoder on train only, then transforming test -- `code` -- categorical columns are encoded with `LabelEncoder`/mapping fitted on the training frame and applied to the test frame	code	mle_clean_v3	da-pool-clean/dsbench-cat-in-the-dat@s3#1	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	10	Fitting a label/ordinal encoder on train only, then transforming test	### Fitting a label/ordinal encoder on train only, then transforming test - Kind: operational - Applies when: `code` -- categorical columns are encoded with `LabelEncoder`/mapping fitted on the training frame and applied to the test frame - Pattern: `le.fit_transform(train[col]); le.transform(test[col])` fo...	30	25	5	39	39	[ "da-pool-clean/dsbench-cat-in-the-dat@s3#1", "da-pool-clean/dsbench-microsoft-malware-prediction@s4#0", "da-pool-clean/dsbench-cat-in-the-dat@s4#0", "da-pool-clean/dsbench-cat-in-the-dat#0", "da-pool-clean/dsbench-cat-in-the-dat@s2#2", "da-pool-clean/dsbench-porto-seguro-safe-driver-prediction@s3#2", "d...
4012d12e137c	### Hyperparameters borrowed from a different library's API - Kind: operational - Applies when: `code` -- an sklearn estimator is constructed with keyword arguments copied from XGBoost/LightGBM idiom. - Pattern: e.g. `GradientBoostingClassifier(..., colsample_bytree=..., reg_lambda=..., early_stopping_round...	Hyperparameters borrowed from a different library's API -- `code` -- an sklearn estimator is constructed with keyword arguments copied from XGBoost/LightGBM idiom.	code	mle_clean_v3	da-pool-clean/dsbench-porto-seguro-safe-driver-prediction@s2#3	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	11	Hyperparameters borrowed from a different library's API	### Hyperparameters borrowed from a different library's API - Kind: operational - Applies when: `code` -- an sklearn estimator is constructed with keyword arguments copied from XGBoost/LightGBM idiom. - Pattern: e.g. `GradientBoostingClassifier(..., colsample_bytree=..., reg_lambda=..., early_stopping_round...	20	20	0	18	18	[ "da-pool-clean/dsbench-porto-seguro-safe-driver-prediction@s2#3", "da-pool-clean/dsbench-cat-in-the-dat@s3#3", "da-pool-clean/dsbench-instant-gratification@s3#1", "da-pool-clean/dsbench-porto-seguro-safe-driver-prediction#2", "da-pool-clean/dsbench-santander-customer-transaction-prediction@s3#1", "da-pool...
db86491552c8	### One-estimator-per-output wrapper over hundreds of targets - Kind: operational - Applies when: `code` -- a multi-output prediction problem is fit with `MultiOutputClassifier`/`MultiOutputRegressor` or a per-column loop - Pattern: wrapping a tree ensemble or iterative linear model in a per-target wrapper ...	One-estimator-per-output wrapper over hundreds of targets -- `code` -- a multi-output prediction problem is fit with `MultiOutputClassifier`/`MultiOutputRegressor` or a per-column loop	code	mle_clean_v3	da-pool-clean/dsbench-conways-reverse-game-of-life-2020@s3#1	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	12	One-estimator-per-output wrapper over hundreds of targets	### One-estimator-per-output wrapper over hundreds of targets - Kind: operational - Applies when: `code` -- a multi-output prediction problem is fit with `MultiOutputClassifier`/`MultiOutputRegressor` or a per-column loop - Pattern: wrapping a tree ensemble or iterative linear model in a per-target wrapper ...	20	20	0	8	8	[ "da-pool-clean/dsbench-conways-reverse-game-of-life-2020@s3#1", "da-pool-clean/dsbench-conways-reverse-game-of-life-2020@s5#2", "da-pool-clean/dsbench-conways-reverse-game-of-life-2020#1", "da-pool-clean/dsbench-conways-reverse-game-of-life-2020@s2#1", "da-pool-clean/dsbench-conways-reverse-game-of-life-202...
d8cfb69cf722	### Rounded p-value not formatted to the requested decimal count - Kind: methodology - Applies when: `task` -- the answer format fixes a number of decimal places for a reported statistic (e.g. four decimals for a p-value) - Pattern: the value is emitted with `round(x, 4)` and Python's default repr, so an un...	Rounded p-value not formatted to the requested decimal count -- `task` -- the answer format fixes a number of decimal places for a reported statistic (e.g. four decimals for a p-value)	task	mle_clean_v3	da-pool-clean-da/dabench-140#1	infiagent-dabench	clean:methodology	mle-rubrics-clean-v3-oraclepool	13	Rounded p-value not formatted to the requested decimal count	### Rounded p-value not formatted to the requested decimal count - Kind: methodology - Applies when: `task` -- the answer format fixes a number of decimal places for a reported statistic (e.g. four decimals for a p-value) - Pattern: the value is emitted with `round(x, 4)` and Python's default repr, so an un...	19	0	19	181	182	[ "da-pool-clean-da/dabench-140#1", "da-pool-clean-da/dabench-268#0", "da-pool-clean-da/dabench-73#1", "da-pool-clean-da/dabench-662#1", "da-pool-clean-da/dabench-647#1", "da-pool-clean-da/dabench-33#1", "da-pool-clean-da/dabench-66#1", "da-pool-clean-da/dabench-244#1", "da-pool-clean-da/dabench-665#4...
c19deca23ff1	### Output schema invented instead of mirrored from the sample file - Kind: methodology - Applies when: `task` -- the task says to write results "in the format specified in" a sample/template file. - Pattern: the script emits a result file with self-chosen column names and extra columns, and/or writes the f...	Output schema invented instead of mirrored from the sample file -- `task` -- the task says to write results "in the format specified in" a sample/template file.	task	mle_clean_v3	da-pool-clean-da/dacode-data-sa-028#2	da-code	clean:methodology	mle-rubrics-clean-v3-oraclepool	14	Output schema invented instead of mirrored from the sample file	### Output schema invented instead of mirrored from the sample file - Kind: methodology - Applies when: `task` -- the task says to write results "in the format specified in" a sample/template file. - Pattern: the script emits a result file with self-chosen column names and extra columns, and/or writes the f...	19	1	18	31	31	[ "da-pool-clean-da/dacode-data-sa-028#2", "da-pool-clean-da/dacode-data-sa-028@s2#4", "da-pool-clean-da/dacode-ml-binary-016@s2#1", "da-pool-clean-da/dacode-data-sa-026@s2#2", "da-pool-clean-da/dacode-ml-regression-008@s3#1", "da-pool-clean-da/dacode-data-sa-028@s3#3", "da-pool-clean-da/dacode-ml-binary-...
f2946bc44ea0	### dtype-string comparison to detect object/text columns - Kind: operational - Applies when: `code` -- column-type branching is done with an equality test against a dtype name literal - Pattern: `if df[col].dtype == 'object':` (or `== 'str'`, `!= 'int64'`) guarding the conversion of text columns to numeric...	dtype-string comparison to detect object/text columns -- `code` -- column-type branching is done with an equality test against a dtype name literal	code	mle_clean_v3	da-pool-clean/dsbench-cat-in-the-dat#1	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	15	dtype-string comparison to detect object/text columns	### dtype-string comparison to detect object/text columns - Kind: operational - Applies when: `code` -- column-type branching is done with an equality test against a dtype name literal - Pattern: `if df[col].dtype == 'object':` (or `== 'str'`, `!= 'int64'`) guarding the conversion of text columns to numeric...	18	18	0	8	8	[ "da-pool-clean/dsbench-cat-in-the-dat#1", "da-pool-clean/dsbench-cat-in-the-dat@s7#0", "da-pool-clean/dsbench-cat-in-the-dat@s8#1", "da-pool-clean-da/dacode-ml-competition-003@s4#1", "da-pool-v3/dsbench-cat-in-the-dat#1", "da-pool-v3/dsbench-cat-in-the-dat@s2#1", "da-pool-v3/dsbench-cat-in-the-dat-ii@s4...
6293f855b67a	### No held-out evaluation or baseline comparison before submitting - Kind: methodology - Applies when: `task` -- a competition metric is stated and instructions require validating on held-out data - Pattern: scripts compute no error metric at all; "validation" consists solely of row counts, NaN counts and ...	No held-out evaluation or baseline comparison before submitting -- `task` -- a competition metric is stated and instructions require validating on held-out data	task	mle_clean_v3	da-pool-clean/dsbench-covid19-global-forecasting-week-3#1	dsbench	clean:methodology	mle-rubrics-clean-v3-oraclepool	16	No held-out evaluation or baseline comparison before submitting	### No held-out evaluation or baseline comparison before submitting - Kind: methodology - Applies when: `task` -- a competition metric is stated and instructions require validating on held-out data - Pattern: scripts compute no error metric at all; "validation" consists solely of row counts, NaN counts and ...	17	0	17	29	29	[ "da-pool-clean/dsbench-covid19-global-forecasting-week-3#1", "da-pool-clean/dsbench-covid19-global-forecasting-week-2@s3#2", "da-pool-clean/dsbench-covid19-global-forecasting-week-1@s5#2", "da-pool-clean/dsbench-cat-in-the-dat-ii@s7#4", "da-pool-clean/dsbench-covid19-global-forecasting-week-4@s5#0", "da-p...
c57c5cf0a186	### Deprecated/removed pandas keyword arguments - Kind: operational - Applies when: `code` -- missing-value or resampling calls use keyword forms removed in recent pandas (e.g. `fillna(method='bfill'/'ffill')`, `append`, `inplace` on removed APIs). - Pattern: Legacy-style API call that the installed library...	Deprecated/removed pandas keyword arguments -- `code` -- missing-value or resampling calls use keyword forms removed in recent pandas (e.g. `fillna(method='bfill'/'ffill')`, `append`, `inplace` on removed APIs).	code	mle_clean_v3	da-pool-clean/dsbench-liverpool-ion-switching#1	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	18	Deprecated/removed pandas keyword arguments	### Deprecated/removed pandas keyword arguments - Kind: operational - Applies when: `code` -- missing-value or resampling calls use keyword forms removed in recent pandas (e.g. `fillna(method='bfill'/'ffill')`, `append`, `inplace` on removed APIs). - Pattern: Legacy-style API call that the installed library...	16	16	0	10	10	[ "da-pool-clean/dsbench-liverpool-ion-switching#1", "da-pool-clean/dsbench-demand-forecasting-kernels-only@s4#1", "da-pool-clean/dsbench-liverpool-ion-switching@s3#0", "da-pool-clean-da/dacode-ml-regression-002@s2#1", "da-pool-clean-da/dacode-ml-regression-002@s3#2", "da-pool-clean-da/dacode-data-sa-029@s2...
829bc24be10d	### Answer string does not reproduce the literal token format (quotes/separators) shown in the task - Kind: methodology - Applies when: `task` -- the task dictates an exact answer template with `@key[value]` tokens where the sample values appear quoted or otherwise decorated. - Pattern: The final answer fil...	Answer string does not reproduce the literal token format (quotes/separators) shown in the task -- `task` -- the task dictates an exact answer template with `@key[value]` tokens where the sample values appear quoted or otherwise decorated.	task	mle_clean_v3	da-pool-clean-da/dabench-550#0	infiagent-dabench	clean:methodology	mle-rubrics-clean-v3-oraclepool	19	Answer string does not reproduce the literal token format (quotes/separators) shown in the task	### Answer string does not reproduce the literal token format (quotes/separators) shown in the task - Kind: methodology - Applies when: `task` -- the task dictates an exact answer template with `@key[value]` tokens where the sample values appear quoted or otherwise decorated. - Pattern: The final answer fil...	15	0	15	18	18	[ "da-pool-clean-da/dabench-550#0", "da-pool-clean-da/dacode-di-text-002@s2#3", "da-pool-clean-da/dabench-550@s2#0", "da-pool-clean-da/dabench-451@s2#1", "da-pool-clean-da/dabench-14@s2#3", "da-pool-clean-da/dacode-di-text-002@s3#2", "da-pool-clean-da/dabench-550@s3#0", "da-pool-clean-da/dabench-550@s4#...
96eaffdb7ad0	### Off-by-one indexing of required output column names - Kind: methodology - Applies when: `task` -- the spec names output columns with an index expression such as "Column_i where i is the ith element" - Pattern: columns generated with `enumerate(...)` starting at 0 (`f'Feature_{i}'` → `Feature_0 ...`) whe...	Off-by-one indexing of required output column names -- `task` -- the spec names output columns with an index expression such as "Column_i where i is the ith element"	task	mle_clean_v3	da-pool-clean-da/dacode-ml-cluster-013#1	da-code	clean:methodology	mle-rubrics-clean-v3-oraclepool	20	Off-by-one indexing of required output column names	### Off-by-one indexing of required output column names - Kind: methodology - Applies when: `task` -- the spec names output columns with an index expression such as "Column_i where i is the ith element" - Pattern: columns generated with `enumerate(...)` starting at 0 (`f'Feature_{i}'` → `Feature_0 ...`) whe...	15	0	15	16	16	[ "da-pool-clean-da/dacode-ml-cluster-013#1", "da-pool-clean-da/dacode-ml-cluster-013@s2#2", "da-pool-clean-da/dacode-ml-cluster-010@s2#4", "da-pool-clean-da/dacode-ml-cluster-019@s2#3", "da-pool-clean-da/dacode-ml-cluster-014@s3#4", "da-pool-clean-da/dacode-ml-cluster-010@s3#2", "da-pool-clean-da/dacode-...
2a6f2d342045	### Text/CSV read with default UTF-8 on data containing non-ASCII bytes - Kind: operational - Applies when: `code` -- reading delimited text files whose content includes names/labels from non-English locales. - Pattern: `pd.read_csv(path)` with no `encoding=` and no error handling, on files that carry accen...	Text/CSV read with default UTF-8 on data containing non-ASCII bytes -- `code` -- reading delimited text files whose content includes names/labels from non-English locales.	code	mle_clean_v3	da-pool-clean-da/dacode-plot-scatter-002#1	da-code	clean:operational	mle-rubrics-clean-v3-oraclepool	21	Text/CSV read with default UTF-8 on data containing non-ASCII bytes	### Text/CSV read with default UTF-8 on data containing non-ASCII bytes - Kind: operational - Applies when: `code` -- reading delimited text files whose content includes names/labels from non-English locales. - Pattern: `pd.read_csv(path)` with no `encoding=` and no error handling, on files that carry accen...	15	15	0	15	15	[ "da-pool-clean-da/dacode-plot-scatter-002#1", "da-pool-clean-da/dacode-ml-multi-011@s2#0", "da-pool-clean-da/dacode-plot-scatter-002@s2#1", "da-pool-clean-da/dacode-ml-cluster-019@s2#0", "da-pool-clean-da/dacode-plot-pie-008@s3#0", "da-pool-clean-da/dacode-plot-pie-008@s4#1", "da-pool-clean-da/dacode-pl...
d64a872fa455	### Mixed-offset timestamp parsing without an explicit UTC flag - Kind: operational - Applies when: `code` -- parsing a string timestamp column that carries UTC offsets (e.g. `...+01:00` / `...+02:00` across a DST boundary). - Pattern: `pd.to_datetime(df[col])` with no `utc=True` (or `format=`/`tz=`) on a c...	Mixed-offset timestamp parsing without an explicit UTC flag -- `code` -- parsing a string timestamp column that carries UTC offsets (e.g. `...+01:00` / `...+02:00` across a DST boundary).	code	mle_clean_v3	da-pool-clean-da/dacode-ml-regression-002@s2#0	da-code	clean:operational	mle-rubrics-clean-v3-oraclepool	24	Mixed-offset timestamp parsing without an explicit UTC flag	### Mixed-offset timestamp parsing without an explicit UTC flag - Kind: operational - Applies when: `code` -- parsing a string timestamp column that carries UTC offsets (e.g. `...+01:00` / `...+02:00` across a DST boundary). - Pattern: `pd.to_datetime(df[col])` with no `utc=True` (or `format=`/`tz=`) on a c...	12	12	0	6	6	[ "da-pool-clean-da/dacode-ml-regression-002@s2#0", "da-pool-clean-da/dacode-ml-regression-002@s4#0", "da-pool-v3-da/dacode-ml-regression-002#0", "da-pool-v3-da/dacode-ml-regression-002@s2#0", "da-pool-v3-da/dacode-ml-regression-002@s3#0", "da-pool-v3-da/dacode-ml-regression-002@s4#0" ]
900fd6b5a305	### Answer file hand-written rather than emitted by the script - Kind: methodology - Applies when: `code` -- the final answer file must contain values computed by the analysis. - Pattern: The script prints results to stdout, and the answer file is then created by a separate shell heredoc with the numbers an...	Answer file hand-written rather than emitted by the script -- `code` -- the final answer file must contain values computed by the analysis.	code	mle_clean_v3	da-pool-clean-da/dabench-33#2	infiagent-dabench	clean:methodology	mle-rubrics-clean-v3-oraclepool	25	Answer file hand-written rather than emitted by the script	### Answer file hand-written rather than emitted by the script - Kind: methodology - Applies when: `code` -- the final answer file must contain values computed by the analysis. - Pattern: The script prints results to stdout, and the answer file is then created by a separate shell heredoc with the numbers an...	11	5	6	57	57	[ "da-pool-clean-da/dabench-33#2", "da-pool-clean-da/dabench-298#4", "da-pool-clean-da/dabench-650#3", "da-pool-clean-da/dabench-71#1", "da-pool-clean-da/dabench-658#1", "da-pool-clean-da/dabench-543#1", "da-pool-clean-da/dabench-465#3", "da-pool-clean-da/dabench-663#2", "da-pool-clean-da/dabench-495@...
03258c0967c0	### Module referenced before (or without) its import - Kind: operational - Applies when: `code` -- a script uses a library alias in a check or print - Pattern: `np.isinf(...)` / similar appears above the `import numpy as np` line, or the import is missing entirely, because the snippet was assembled incremen...	Module referenced before (or without) its import -- `code` -- a script uses a library alias in a check or print	code	mle_clean_v3	da-pool-clean/dsbench-covid19-global-forecasting-week-1@s2#1	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	26	Module referenced before (or without) its import	### Module referenced before (or without) its import - Kind: operational - Applies when: `code` -- a script uses a library alias in a check or print - Pattern: `np.isinf(...)` / similar appears above the `import numpy as np` line, or the import is missing entirely, because the snippet was assembled incremen...	11	11	0	11	11	[ "da-pool-clean/dsbench-covid19-global-forecasting-week-1@s2#1", "da-pool-clean/dsbench-covid19-global-forecasting-week-2@s4#0", "da-pool-clean/dsbench-santander-customer-satisfaction@s5#0", "da-pool-v3/dsbench-bike-sharing-demand@s3#1", "da-pool-v3/dsbench-covid19-global-forecasting-week-4@s6#0", "da-pool...
55d2ae5d6f80	### Assuming test dates are strictly after training dates without checking - Kind: methodology - Applies when: `task` -- a forecasting task whose split may be by row rather than by time, or whose test file shares the training date range - Pattern: the first models compute a "last observed value" per group a...	Assuming test dates are strictly after training dates without checking -- `task` -- a forecasting task whose split may be by row rather than by time, or whose test file shares the training date range	task	mle_clean_v3	da-pool-clean/dsbench-covid19-global-forecasting-week-1#4	dsbench	clean:methodology	mle-rubrics-clean-v3-oraclepool	27	Assuming test dates are strictly after training dates without checking	### Assuming test dates are strictly after training dates without checking - Kind: methodology - Applies when: `task` -- a forecasting task whose split may be by row rather than by time, or whose test file shares the training date range - Pattern: the first models compute a "last observed value" per group a...	10	0	10	14	14	[ "da-pool-clean/dsbench-covid19-global-forecasting-week-1#4", "da-pool-clean/dsbench-covid19-global-forecasting-week-2@s5#1", "da-pool-clean/dsbench-covid19-global-forecasting-week-4@s5#2", "da-pool-clean/dsbench-covid19-global-forecasting-week-3@s6#0", "da-pool-v3/dsbench-covid19-global-forecasting-week-1@s...
9aea3b5c1c95	### Deprecated/removed pandas frequency and API aliases - Kind: operational - Applies when: `code` -- time-series code passes legacy offset aliases or uses APIs removed in the installed major version. - Pattern: Calls like `date_range(..., freq='M')` (also `'H'`, `'T'`, `append`, `.iteritems`) written again...	Deprecated/removed pandas frequency and API aliases -- `code` -- time-series code passes legacy offset aliases or uses APIs removed in the installed major version.	code	mle_clean_v3	da-pool-clean-da/dacode-plot-line-015@s2#1	da-code	clean:operational	mle-rubrics-clean-v3-oraclepool	28	Deprecated/removed pandas frequency and API aliases	### Deprecated/removed pandas frequency and API aliases - Kind: operational - Applies when: `code` -- time-series code passes legacy offset aliases or uses APIs removed in the installed major version. - Pattern: Calls like `date_range(..., freq='M')` (also `'H'`, `'T'`, `append`, `.iteritems`) written again...	10	10	0	10	10	[ "da-pool-clean-da/dacode-plot-line-015@s2#1", "da-pool-clean-da/dacode-data-sa-043@s2#0", "da-pool-clean-da/dacode-data-sa-043@s3#0", "da-pool-v3-da/dacode-data-sa-043#0", "da-pool-v3-da/dacode-plot-line-015@s2#1", "da-pool-v3-da/dacode-data-sa-043@s2#0", "da-pool-v3-da/dacode-ml-regression-002@s2#1", ...
6195417e3e41	### Passing a constructor argument the installed sklearn version removed - Kind: operational - Applies when: `code` -- estimators are constructed with keyword arguments that were deprecated in older sklearn releases. - Pattern: `LogisticRegression(..., multi_class='multinomial', ...)` (same class of issue: ...	Passing a constructor argument the installed sklearn version removed -- `code` -- estimators are constructed with keyword arguments that were deprecated in older sklearn releases.	code	mle_clean_v3	da-pool-clean/dsbench-liverpool-ion-switching@s2#3	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	29	Passing a constructor argument the installed sklearn version removed	### Passing a constructor argument the installed sklearn version removed - Kind: operational - Applies when: `code` -- estimators are constructed with keyword arguments that were deprecated in older sklearn releases. - Pattern: `LogisticRegression(..., multi_class='multinomial', ...)` (same class of issue: ...	10	10	0	9	9	[ "da-pool-clean/dsbench-liverpool-ion-switching@s2#3", "da-pool-clean-da/dacode-ml-multi-011#0", "da-pool-clean-da/dacode-ml-competition-005@s2#0", "da-pool-clean-da/dacode-ml-competition-006@s3#1", "da-pool-v3/dsbench-liverpool-ion-switching@s3#1", "da-pool-v3/dsbench-porto-seguro-safe-driver-prediction@s...
e7f887530dd4	### Row-by-row DataFrame filtering inside a loop over the test set - Kind: operational - Applies when: `code` -- predictions are produced by iterating with `iterrows()` and re-filtering the full training frame for each row - Pattern: `for _, row in test.iterrows(): hist = train[train[key] == row[key]].sort_...	Row-by-row DataFrame filtering inside a loop over the test set -- `code` -- predictions are produced by iterating with `iterrows()` and re-filtering the full training frame for each row	code	mle_clean_v3	da-pool-clean/dsbench-covid19-global-forecasting-week-3#4	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	30	Row-by-row DataFrame filtering inside a loop over the test set	### Row-by-row DataFrame filtering inside a loop over the test set - Kind: operational - Applies when: `code` -- predictions are produced by iterating with `iterrows()` and re-filtering the full training frame for each row - Pattern: `for _, row in test.iterrows(): hist = train[train[key] == row[key]].sort_...	9	3	6	34	34	[ "da-pool-clean/dsbench-covid19-global-forecasting-week-3#4", "da-pool-clean/dsbench-covid19-global-forecasting-week-3@s3#1", "da-pool-clean/dsbench-covid19-global-forecasting-week-1#1", "da-pool-clean/dsbench-covid19-global-forecasting-week-1@s4#5", "da-pool-clean/dsbench-covid19-global-forecasting-week-3@s...
9b37a8577928	### Imported symbol that the library never exported - Kind: operational - Applies when: `code` -- a script imports a named metric/helper from a library (e.g. `from sklearn.metrics import <competition_metric_name>`). - Pattern: The competition metric's colloquial name is imported verbatim from a package that...	Imported symbol that the library never exported -- `code` -- a script imports a named metric/helper from a library (e.g. `from sklearn.metrics import <competition_metric_name>`).	code	mle_clean_v3	da-pool-clean-da/dacode-ml-competition-006#0	da-code	clean:operational	mle-rubrics-clean-v3-oraclepool	31	Imported symbol that the library never exported	### Imported symbol that the library never exported - Kind: operational - Applies when: `code` -- a script imports a named metric/helper from a library (e.g. `from sklearn.metrics import <competition_metric_name>`). - Pattern: The competition metric's colloquial name is imported verbatim from a package that...	9	9	0	8	8	[ "da-pool-clean-da/dacode-ml-competition-006#0", "da-pool-clean-da/dacode-ml-competition-003@s2#0", "da-pool-clean-da/dacode-ml-competition-006@s2#0", "da-pool-clean-da/dacode-ml-competition-006@s3#0", "da-pool-clean-da/dacode-ml-competition-006@s4#0", "da-pool-v3-da/dacode-ml-competition-003@s2#0", "da-...
84ee28a8727e	### Required reference/spec file absent, mapping invented - Kind: causal - Applies when: `task` -- the prompt tells the agent to use a definition, label mapping, or rule set contained in a named auxiliary file - Pattern: The named file is not found, and the code proceeds with a hard-coded mapping taken from...	Required reference/spec file absent, mapping invented -- `task` -- the prompt tells the agent to use a definition, label mapping, or rule set contained in a named auxiliary file	task	mle_clean_v3	da-pool-clean-da/dacode-di-text-004#1	da-code	clean:causal	mle-rubrics-clean-v3-oraclepool	32	Required reference/spec file absent, mapping invented	### Required reference/spec file absent, mapping invented - Kind: causal - Applies when: `task` -- the prompt tells the agent to use a definition, label mapping, or rule set contained in a named auxiliary file - Pattern: The named file is not found, and the code proceeds with a hard-coded mapping taken from...	9	2	7	7	7	[ "da-pool-clean-da/dacode-di-text-004#1", "da-pool-clean-da/dacode-di-text-004@s2#0", "da-pool-clean-da/dacode-di-text-004@s3#0", "da-pool-v3-da/dacode-di-text-004#0", "da-pool-v3-da/dacode-di-text-004@s2#0", "da-pool-v3-da/dacode-di-text-004@s3#0", "da-pool-v3-da/dacode-di-text-004@s4#0" ]
29d6dff4f491	### Deliverable replaced by a prose narrative - Kind: methodology - Applies when: `task` -- instructions demand analysis kept in saved `.py` files and a final answer in an exact format - Pattern: all computation is done through inline heredoc/stdin invocations so no script persists, and the final answer fil...	Deliverable replaced by a prose narrative -- `task` -- instructions demand analysis kept in saved `.py` files and a final answer in an exact format	task	mle_clean_v3	da-pool-clean-da/dacode-ml-regression-008#5	da-code	clean:methodology	mle-rubrics-clean-v3-oraclepool	33	Deliverable replaced by a prose narrative	### Deliverable replaced by a prose narrative - Kind: methodology - Applies when: `task` -- instructions demand analysis kept in saved `.py` files and a final answer in an exact format - Pattern: all computation is done through inline heredoc/stdin invocations so no script persists, and the final answer fil...	9	5	4	4	4	[ "da-pool-clean-da/dacode-ml-regression-008#5", "da-pool-clean-da/dacode-ml-cluster-019@s3#5", "da-pool-v3-da/dacode-ml-regression-008@s3#5", "da-pool-v3-da/dacode-plot-line-015@s4#5" ]
373d3b47fff8	### Per-element `LabelEncoder.transform` inside a row-wise `map`/`apply` - Kind: operational - Applies when: `code` -- categorical columns are encoded by calling a fitted encoder once per cell - Pattern: `df[col].map(lambda x: le.transform([x])[0] if pd.notna(x) else -1)` (or `apply`) executed for every col...	Per-element `LabelEncoder.transform` inside a row-wise `map`/`apply` -- `code` -- categorical columns are encoded by calling a fitted encoder once per cell	code	mle_clean_v3	da-pool-clean/dsbench-microsoft-malware-prediction#0	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	34	Per-element `LabelEncoder.transform` inside a row-wise `map`/`apply`	### Per-element `LabelEncoder.transform` inside a row-wise `map`/`apply` - Kind: operational - Applies when: `code` -- categorical columns are encoded by calling a fitted encoder once per cell - Pattern: `df[col].map(lambda x: le.transform([x])[0] if pd.notna(x) else -1)` (or `apply`) executed for every col...	9	9	0	3	3	[ "da-pool-clean/dsbench-microsoft-malware-prediction#0", "da-pool-v3/dsbench-microsoft-malware-prediction@s5#0", "da-pool-v3/dsbench-microsoft-malware-prediction@s3#1" ]
5bf7cb614df6	### Categorical columns detected with `dtype == 'object'` - Kind: operational - Applies when: `code` -- encoding text columns of a CSV read with a modern pandas where string columns get a dedicated `str`/`string` dtype rather than `object` - Pattern: the categorical/numeric split is decided by `df[col].dtyp...	Categorical columns detected with `dtype == 'object'` -- `code` -- encoding text columns of a CSV read with a modern pandas where string columns get a dedicated `str`/`string` dtype rather than `object`	code	mle_clean_v3	da-pool-v3/dsbench-microsoft-malware-prediction@s7#0	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	35	Categorical columns detected with `dtype == 'object'`	### Categorical columns detected with `dtype == 'object'` - Kind: operational - Applies when: `code` -- encoding text columns of a CSV read with a modern pandas where string columns get a dedicated `str`/`string` dtype rather than `object` - Pattern: the categorical/numeric split is decided by `df[col].dtyp...	9	9	0	1	1	[ "da-pool-v3/dsbench-microsoft-malware-prediction@s7#0" ]
a84ec057c15f	### Squared-error objective under a log-scale metric - Kind: methodology - Applies when: `task` -- the metric is a logarithmic/relative error (e.g. RMSLE) or otherwise not plain RMSE on the raw target - Pattern: regressors are fit directly on the raw skewed count target with default squared-error loss and p...	Squared-error objective under a log-scale metric -- `task` -- the metric is a logarithmic/relative error (e.g. RMSLE) or otherwise not plain RMSE on the raw target	task	mle_clean_v3	da-pool-clean/dsbench-bike-sharing-demand@s2#1	dsbench	clean:methodology	mle-rubrics-clean-v3-oraclepool	37	Squared-error objective under a log-scale metric	### Squared-error objective under a log-scale metric - Kind: methodology - Applies when: `task` -- the metric is a logarithmic/relative error (e.g. RMSLE) or otherwise not plain RMSE on the raw target - Pattern: regressors are fit directly on the raw skewed count target with default squared-error loss and p...	8	1	7	32	32	[ "da-pool-clean/dsbench-bike-sharing-demand@s2#1", "da-pool-clean/dsbench-bike-sharing-demand#1", "da-pool-clean/dsbench-bike-sharing-demand@s3#0", "da-pool-clean/dsbench-bike-sharing-demand@s4#1", "da-pool-clean/dsbench-santander-value-prediction-challenge#2", "da-pool-clean/dsbench-bike-sharing-demand@s5...
01ce2e69b182	### Task-mandated config/spec file never located before producing the artifact - Kind: methodology - Applies when: `task` -- the task says the output must be formatted according to a named spec file (yaml/json/config) - Pattern: The agent checks one directory, does not find the spec, and then invents its ow...	Task-mandated config/spec file never located before producing the artifact -- `task` -- the task says the output must be formatted according to a named spec file (yaml/json/config)	task	mle_clean_v3	da-pool-clean-da/dacode-plot-line-006#1	da-code	clean:methodology	mle-rubrics-clean-v3-oraclepool	38	Task-mandated config/spec file never located before producing the artifact	### Task-mandated config/spec file never located before producing the artifact - Kind: methodology - Applies when: `task` -- the task says the output must be formatted according to a named spec file (yaml/json/config) - Pattern: The agent checks one directory, does not find the spec, and then invents its ow...	8	4	4	13	13	[ "da-pool-clean-da/dacode-plot-line-006#1", "da-pool-clean-da/dacode-plot-bar-007#3", "da-pool-clean-da/dacode-plot-line-006@s2#1", "da-pool-clean-da/dacode-plot-line-006@s3#2", "da-pool-clean-da/dacode-plot-bar-007@s3#3", "da-pool-clean-da/dacode-plot-line-006@s4#1", "da-pool-clean-da/dacode-plot-bar-00...
025513d64dc7	### Output column format guessed rather than derived from the template - Kind: methodology - Applies when: `task` -- a sample/template output file is provided with headers but blank value cells - Pattern: the script invents a value encoding (e.g. packing two numbers into one bracketed string, fixed decimal ...	Output column format guessed rather than derived from the template -- `task` -- a sample/template output file is provided with headers but blank value cells	task	mle_clean_v3	da-pool-clean-da/dacode-data-sa-029#3	da-code	clean:methodology	mle-rubrics-clean-v3-oraclepool	39	Output column format guessed rather than derived from the template	### Output column format guessed rather than derived from the template - Kind: methodology - Applies when: `task` -- a sample/template output file is provided with headers but blank value cells - Pattern: the script invents a value encoding (e.g. packing two numbers into one bracketed string, fixed decimal ...	8	0	8	8	8	[ "da-pool-clean-da/dacode-data-sa-029#3", "da-pool-clean-da/dacode-dm-csv-015@s2#1", "da-pool-clean-da/dacode-dm-csv-015@s3#0", "da-pool-clean-da/dacode-dm-csv-015#0", "da-pool-v3-da/dacode-dm-csv-015#0", "da-pool-v3-da/dacode-dm-csv-015@s2#0", "da-pool-v3-da/dacode-dm-csv-015@s4#0", "da-pool-v3-da/dac...
74919fe9d57e	### Missing-value and non-numeric handling not verified before a moment statistic - Kind: methodology - Applies when: `code` -- a numeric statistic is computed on a column of a CSV that may contain sentinel strings (`-`, `N/A`, footnote marks) or blanks - Pattern: computing `skew`/`mean` directly on a parse...	Missing-value and non-numeric handling not verified before a moment statistic -- `code` -- a numeric statistic is computed on a column of a CSV that may contain sentinel strings (`-`, `N/A`, footnote marks) or blanks	code	mle_clean_v3	da-pool-clean-da/dabench-19#2	infiagent-dabench	clean:methodology	mle-rubrics-clean-v3-oraclepool	40	Missing-value and non-numeric handling not verified before a moment statistic	### Missing-value and non-numeric handling not verified before a moment statistic - Kind: methodology - Applies when: `code` -- a numeric statistic is computed on a column of a CSV that may contain sentinel strings (`-`, `N/A`, footnote marks) or blanks - Pattern: computing `skew`/`mean` directly on a parse...	7	2	5	99	99	[ "da-pool-clean-da/dabench-19#2", "da-pool-clean-da/dabench-724#3", "da-pool-clean-da/dabench-375#2", "da-pool-clean-da/dabench-656#1", "da-pool-clean-da/dabench-423#1", "da-pool-clean-da/dabench-18#1", "da-pool-clean-da/dabench-297#2", "da-pool-clean-da/dabench-55#0", "da-pool-clean-da/dabench-419@s...
9b569e04d037	### Parametric two-sample t-test applied to skewed discrete count data - Kind: methodology - Applies when: `code` -- the script compares the means of two samples of counts/bounded integers and reaches for `scipy.stats.ttest_ind` without any distributional check. - Pattern: `t_stat, p_val = stats.ttest_ind(a...	Parametric two-sample t-test applied to skewed discrete count data -- `code` -- the script compares the means of two samples of counts/bounded integers and reaches for `scipy.stats.ttest_ind` without any distributional check.	code	mle_clean_v3	da-pool-clean-da/dacode-data-sa-001#0	da-code	clean:methodology	mle-rubrics-clean-v3-oraclepool	44	Parametric two-sample t-test applied to skewed discrete count data	### Parametric two-sample t-test applied to skewed discrete count data - Kind: methodology - Applies when: `code` -- the script compares the means of two samples of counts/bounded integers and reaches for `scipy.stats.ttest_ind` without any distributional check. - Pattern: `t_stat, p_val = stats.ttest_ind(a...	7	0	7	8	8	[ "da-pool-clean-da/dacode-data-sa-001#0", "da-pool-clean-da/dacode-data-sa-001@s2#0", "da-pool-clean-da/dacode-data-sa-001@s3#1", "da-pool-clean-da/dacode-data-sa-001@s4#2", "da-pool-v3-da/dacode-data-sa-001#2", "da-pool-v3-da/dacode-data-sa-001@s2#1", "da-pool-v3-da/dacode-data-sa-001@s3#0", "da-pool-...
523f99a4b1a7	### `.days` attribute on numpy datetime/timedelta values - Kind: operational - Applies when: `code` -- date arithmetic is done after pulling a datetime column out of a DataFrame with `.values` / `to_numpy()` - Pattern: converting a column to a numpy array (dtype `datetime64[ns]`) and then subtracting two el...	`.days` attribute on numpy datetime/timedelta values -- `code` -- date arithmetic is done after pulling a datetime column out of a DataFrame with `.values` / `to_numpy()`	code	mle_clean_v3	da-pool-clean/dsbench-covid19-global-forecasting-week-2#0	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	49	`.days` attribute on numpy datetime/timedelta values	### `.days` attribute on numpy datetime/timedelta values - Kind: operational - Applies when: `code` -- date arithmetic is done after pulling a datetime column out of a DataFrame with `.values` / `to_numpy()` - Pattern: converting a column to a numpy array (dtype `datetime64[ns]`) and then subtracting two el...	7	7	0	5	5	[ "da-pool-clean/dsbench-covid19-global-forecasting-week-2#0", "da-pool-clean/dsbench-covid19-global-forecasting-week-3@s2#0", "da-pool-clean/dsbench-covid19-global-forecasting-week-4@s8#0", "da-pool-v3/dsbench-covid19-global-forecasting-week-1@s2#0", "da-pool-v3/dsbench-covid19-global-forecasting-week-4@s3#0...
bd5a0a9928a9	### Imputation statistics computed from an already-imputed frame and applied to test - Kind: methodology - Applies when: `code` -- the script fills training NaNs in place and then reuses the same frame to compute fill values for the test frame - Pattern: `X = X.fillna(X.median())` followed by `X_test = X_te...	Imputation statistics computed from an already-imputed frame and applied to test -- `code` -- the script fills training NaNs in place and then reuses the same frame to compute fill values for the test frame	code	mle_clean_v3	da-pool-clean/dsbench-santander-customer-satisfaction#4	dsbench	clean:methodology	mle-rubrics-clean-v3-oraclepool	50	Imputation statistics computed from an already-imputed frame and applied to test	### Imputation statistics computed from an already-imputed frame and applied to test - Kind: methodology - Applies when: `code` -- the script fills training NaNs in place and then reuses the same frame to compute fill values for the test frame - Pattern: `X = X.fillna(X.median())` followed by `X_test = X_te...	6	2	4	14	14	[ "da-pool-clean/dsbench-santander-customer-satisfaction#4", "da-pool-clean/dsbench-microsoft-malware-prediction@s2#3", "da-pool-clean/dsbench-commonlitreadabilityprize@s4#0", "da-pool-clean/dsbench-tmdb-box-office-prediction@s4#3", "da-pool-clean/dsbench-santander-customer-satisfaction@s7#3", "da-pool-clea...
8f6d24670b33	### NaN values fed into `statistics` module functions - Kind: operational - Applies when: `code` -- a derived numeric column is converted to a Python list and passed to `statistics.stdev`/`mean`/`variance`. - Pattern: `series.tolist()` (or `list(series)`) handed straight to a `statistics` function without f...	NaN values fed into `statistics` module functions -- `code` -- a derived numeric column is converted to a Python list and passed to `statistics.stdev`/`mean`/`variance`.	code	mle_clean_v3	da-pool-clean-da/dabench-662#0	infiagent-dabench	clean:operational	mle-rubrics-clean-v3-oraclepool	52	NaN values fed into `statistics` module functions	### NaN values fed into `statistics` module functions - Kind: operational - Applies when: `code` -- a derived numeric column is converted to a Python list and passed to `statistics.stdev`/`mean`/`variance`. - Pattern: `series.tolist()` (or `list(series)`) handed straight to a `statistics` function without f...	6	6	0	7	7	[ "da-pool-clean-da/dabench-662#0", "da-pool-clean-da/dabench-662@s4#0", "da-pool-v3-da/dabench-662#0", "da-pool-v3-da/dabench-662@s2#0", "da-pool-v3-da/dabench-662@s3#0", "da-pool-v3-da/dabench-372@s3#1", "da-pool-v3-da/dabench-662@s4#0" ]
5e07c9ec890c	### Deliverable stores transformed values instead of the dataset's feature vector - Kind: methodology - Applies when: `task` requires an output table of `Feature_i` columns plus a label, and `code` builds it from a preprocessing artifact - Pattern: the saved output frame is constructed from the scaled/encod...	Deliverable stores transformed values instead of the dataset's feature vector -- `task` requires an output table of `Feature_i` columns plus a label, and `code` builds it from a preprocessing artifact	task	mle_clean_v3	da-pool-clean-da/dacode-ml-cluster-014#1	da-code	clean:methodology	mle-rubrics-clean-v3-oraclepool	53	Deliverable stores transformed values instead of the dataset's feature vector	### Deliverable stores transformed values instead of the dataset's feature vector - Kind: methodology - Applies when: `task` requires an output table of `Feature_i` columns plus a label, and `code` builds it from a preprocessing artifact - Pattern: the saved output frame is constructed from the scaled/encod...	6	0	6	6	6	[ "da-pool-clean-da/dacode-ml-cluster-014#1", "da-pool-clean-da/dacode-ml-cluster-013@s2#1", "da-pool-clean-da/dacode-ml-cluster-010@s3#1", "da-pool-clean-da/dacode-ml-cluster-014@s4#1", "da-pool-v3-da/dacode-ml-cluster-013@s2#0", "da-pool-v3-da/dacode-ml-cluster-014@s4#1" ]
05e7beaa920a	### Multi-output target passed to a single-output estimator - Kind: operational - Applies when: `code` -- the label is a 2-D array (one row per sample, many columns) and a scalar-output sklearn estimator is fitted on it - Pattern: `GradientBoostingRegressor(...).fit(X, Y)` / `SVR().fit(X, Y)` with `Y.shape ...	Multi-output target passed to a single-output estimator -- `code` -- the label is a 2-D array (one row per sample, many columns) and a scalar-output sklearn estimator is fitted on it	code	mle_clean_v3	da-pool-clean/dsbench-conways-reverse-game-of-life-2020@s5#1	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	55	Multi-output target passed to a single-output estimator	### Multi-output target passed to a single-output estimator - Kind: operational - Applies when: `code` -- the label is a 2-D array (one row per sample, many columns) and a scalar-output sklearn estimator is fitted on it - Pattern: `GradientBoostingRegressor(...).fit(X, Y)` / `SVR().fit(X, Y)` with `Y.shape ...	6	6	0	5	5	[ "da-pool-clean/dsbench-conways-reverse-game-of-life-2020@s5#1", "da-pool-clean/dsbench-conways-reverse-game-of-life-2020@s8#2", "da-pool-v3/dsbench-conways-reverse-game-of-life-2020#0", "da-pool-v3/dsbench-conways-reverse-game-of-life-2020@s2#3", "da-pool-v3/dsbench-conways-reverse-game-of-life-2020@s8#1" ]
fad5366cdc0f	### Repeated timeouts driving blind feature/model stripping instead of budgeting - Kind: causal - Applies when: `code` -- successive script versions in the same attempt keep the same monolithic "load all → featurize all → fit → predict → write" shape while only deleting features or lowering estimator counts. - ...	Repeated timeouts driving blind feature/model stripping instead of budgeting -- `code` -- successive script versions in the same attempt keep the same monolithic "load all → featurize all → fit → predict → write" shape while only deleting features or lowering estimator counts.	code	mle_clean_v3	da-pool-clean/dsbench-liverpool-ion-switching@s5#5	dsbench	clean:causal	mle-rubrics-clean-v3-oraclepool	56	Repeated timeouts driving blind feature/model stripping instead of budgeting	### Repeated timeouts driving blind feature/model stripping instead of budgeting - Kind: causal - Applies when: `code` -- successive script versions in the same attempt keep the same monolithic "load all → featurize all → fit → predict → write" shape while only deleting features or lowering estimator counts. - ...	6	6	0	1	1	[ "da-pool-clean/dsbench-liverpool-ion-switching@s5#5" ]
c171778620e8	### Per-output-column model loops and pure-Python cell loops that cannot fit the command time limit - Kind: operational - Applies when: `code` -- the target is a wide multi-output array (hundreds of columns) or a grid simulation is needed for many samples, under a fixed per-command timeout. - Pattern: (a) a...	Per-output-column model loops and pure-Python cell loops that cannot fit the command time limit -- `code` -- the target is a wide multi-output array (hundreds of columns) or a grid simulation is needed for many samples, under a fixed per-command timeout.	code	mle_clean_v3	da-pool-clean/dsbench-conways-reverse-game-of-life-2020@s6#3	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	57	Per-output-column model loops and pure-Python cell loops that cannot fit the command time limit	### Per-output-column model loops and pure-Python cell loops that cannot fit the command time limit - Kind: operational - Applies when: `code` -- the target is a wide multi-output array (hundreds of columns) or a grid simulation is needed for many samples, under a fixed per-command timeout. - Pattern: (a) a...	6	6	0	1	1	[ "da-pool-clean/dsbench-conways-reverse-game-of-life-2020@s6#3" ]
4ba57fb4c051	### Declaring completion without verifying the required artifact exists - Kind: causal - Applies when: `code` -- the final submit command is issued after runs that never demonstrably wrote the output file. - Pattern: the memory/time failure in the modeling step kills the write, and the agent nonetheless iss...	Declaring completion without verifying the required artifact exists -- `code` -- the final submit command is issued after runs that never demonstrably wrote the output file.	code	mle_clean_v3	da-pool-v3/dsbench-cat-in-the-dat-ii@s2#5	dsbench	clean:causal	mle-rubrics-clean-v3-oraclepool	58	Declaring completion without verifying the required artifact exists	### Declaring completion without verifying the required artifact exists - Kind: causal - Applies when: `code` -- the final submit command is issued after runs that never demonstrably wrote the output file. - Pattern: the memory/time failure in the modeling step kills the write, and the agent nonetheless iss...	6	5	1	1	1	[ "da-pool-v3/dsbench-cat-in-the-dat-ii@s2#5" ]
9c26ef693b35	### Missing values not provably removed before a NaN-intolerant estimator - Kind: operational - Applies when: `code` -- data with many partially-missing columns is fed to sklearn estimators other than `HistGradientBoosting` - Pattern*: imputation done column-by-column inside a loop that runs before/while co...	Missing values not provably removed before a NaN-intolerant estimator -- `code` -- data with many partially-missing columns is fed to sklearn estimators other than `HistGradientBoosting*`	code	mle_clean_v3	da-pool-v3/dsbench-microsoft-malware-prediction@s7#3	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	59	Missing values not provably removed before a NaN-intolerant estimator	### Missing values not provably removed before a NaN-intolerant estimator - Kind: operational - Applies when: `code` -- data with many partially-missing columns is fed to sklearn estimators other than `HistGradientBoosting` - Pattern*: imputation done column-by-column inside a loop that runs before/while co...	6	6	0	1	1	[ "da-pool-v3/dsbench-microsoft-malware-prediction@s7#3" ]
13a226b43c27	### Per-target model loop whose total fit count cannot finish inside the command timeout - Kind: operational - Applies when: `code` -- a script trains one estimator per output cell/column in a Python loop over hundreds of targets on tens of thousands of rows with hundreds of dense features. - Pattern: `for ...	Per-target model loop whose total fit count cannot finish inside the command timeout -- `code` -- a script trains one estimator per output cell/column in a Python loop over hundreds of targets on tens of thousands of rows with hundreds of dense features.	code	mle_clean_v3	da-pool-v3/dsbench-conways-reverse-game-of-life-2020@s5#1	dsbench	clean:operational	mle-rubrics-clean-v3-oraclepool	60	Per-target model loop whose total fit count cannot finish inside the command timeout	### Per-target model loop whose total fit count cannot finish inside the command timeout - Kind: operational - Applies when: `code` -- a script trains one estimator per output cell/column in a Python loop over hundreds of targets on tens of thousands of rows with hundreds of dense features. - Pattern: `for ...	6	6	0	1	1	[ "da-pool-v3/dsbench-conways-reverse-game-of-life-2020@s5#1" ]
be1b43f7e2f1	### Imputation scope narrowed to the single column of interest without stating it is sufficient - Kind: methodology - Applies when: `task` -- the instruction says to fill missing values with the mean before computing a specific statistic. - Pattern: The script imputes only the one column it will rank, and n...	Imputation scope narrowed to the single column of interest without stating it is sufficient -- `task` -- the instruction says to fill missing values with the mean before computing a specific statistic.	task	mle_clean_v3	da-pool-clean-da/dacode-di-text-002#4	da-code	clean:methodology	mle-rubrics-clean-v3-oraclepool	62	Imputation scope narrowed to the single column of interest without stating it is sufficient	### Imputation scope narrowed to the single column of interest without stating it is sufficient - Kind: methodology - Applies when: `task` -- the instruction says to fill missing values with the mean before computing a specific statistic. - Pattern: The script imputes only the one column it will rank, and n...	5	0	5	25	25	[ "da-pool-clean-da/dacode-di-text-002#4", "da-pool-clean-da/dacode-di-text-003#3", "da-pool-clean-da/dacode-di-text-003@s2#3", "da-pool-clean-da/dabench-453@s2#3", "da-pool-clean-da/dabench-275@s2#4", "da-pool-clean-da/dacode-di-text-002@s3#3", "da-pool-clean-da/dabench-111@s3#3", "da-pool-clean-da/dab...

End of preview. Expand in Data Studio

mle-rubrics-filtered-final

Curation method: Filtered · retrieval used with it: Similarity · rubrics: 220

MLE-bench Lite. The whole-rollout library of the 2nd+3rd collections cut to the collection pool's failure proportions (Sonnet 5.5 labels); the scarcest category sets the size. 220 rubrics.

Rubrics for a coding world model (Claude Opus 5) that reads an agent's code change against 12 retrieved rubrics and predicts whether it works, in place of running it. Mined only from collection-pool rollouts of the same Haiku 4.5 agent on other benchmarks (InfiAgent-DABench, DA-Code, DSBench, MLGym, MLAgentBench, RE-Bench; SUPER for PaperBench), never from the evaluation benchmark except where marked reference. Results, method names and distribution-mismatch plots: https://codingwm.github.io/rubric_distribution.html. Code: https://github.com/7peng/agent-cwm (mining/mine_quota.py, mining/curate_complete.py, mining/mine_clean.py).

Downloads last month: 40

Collection including EdwardoSunny/mle-rubrics-filtered-final

Coding world model: rubric libraries

Collection

Rubric libraries for a coding world model, by curation method; see https://codingwm.github.io/rubric_distribution.html • 12 items • Updated 9 days ago