Dataset Preview
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed
Error code: DatasetGenerationError
Exception: CastError
Message: Couldn't cast
rubric_id: string
family: string
state_id: string
trajectory: string
competition: string
fires: bool
judged: bool
evidence: string
judge_version: string
judge_model: string
eligible_error: bool
static_task_manifest: string
model: string
execution: struct<returncode: int64, timed_out: bool, exception_class: string, exception_message: string, stdou (... 110 chars omitted)
child 0, returncode: int64
child 1, timed_out: bool
child 2, exception_class: string
child 3, exception_message: string
child 4, stdout: string
child 5, stderr: string
child 6, submission_exists: bool
child 7, clean: bool
child 8, exit_ok_no_submission: bool
child 9, seconds: double
code: string
split: string
parent_state: string
task_prompt: string
stage: string
code_sha: string
raw_response: string
to
{'competition': Value('string'), 'trajectory': Value('string'), 'stage': Value('string'), 'state_id': Value('string'), 'parent_state': Value('string'), 'task_prompt': Value('string'), 'raw_response': Value('string'), 'code': Value('string'), 'static_task_manifest': Value('string'), 'code_sha': Value('string'), 'execution': {'returncode': Value('int64'), 'timed_out': Value('bool'), 'exception_class': Value('string'), 'exception_message': Value('string'), 'stdout': Value('string'), 'stderr': Value('string'), 'submission_exists': Value('bool'), 'clean': Value('bool'), 'exit_ok_no_submission': Value('bool'), 'seconds': Value('float64')}, 'eligible_error': Value('bool'), 'model': Value('string'), 'split': Value('string')}
because column names don't match
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1827, in _prepare_split_single
for key, table in generator:
^^^^^^^^^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 613, in wrapped
for item in generator(*args, **kwargs):
~~~~~~~~~^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_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
rubric_id: string
family: string
state_id: string
trajectory: string
competition: string
fires: bool
judged: bool
evidence: string
judge_version: string
judge_model: string
eligible_error: bool
static_task_manifest: string
model: string
execution: struct<returncode: int64, timed_out: bool, exception_class: string, exception_message: string, stdou (... 110 chars omitted)
child 0, returncode: int64
child 1, timed_out: bool
child 2, exception_class: string
child 3, exception_message: string
child 4, stdout: string
child 5, stderr: string
child 6, submission_exists: bool
child 7, clean: bool
child 8, exit_ok_no_submission: bool
child 9, seconds: double
code: string
split: string
parent_state: string
task_prompt: string
stage: string
code_sha: string
raw_response: string
to
{'competition': Value('string'), 'trajectory': Value('string'), 'stage': Value('string'), 'state_id': Value('string'), 'parent_state': Value('string'), 'task_prompt': Value('string'), 'raw_response': Value('string'), 'code': Value('string'), 'static_task_manifest': Value('string'), 'code_sha': Value('string'), 'execution': {'returncode': Value('int64'), 'timed_out': Value('bool'), 'exception_class': Value('string'), 'exception_message': Value('string'), 'stdout': Value('string'), 'stderr': Value('string'), 'submission_exists': Value('bool'), 'clean': Value('bool'), 'exit_ok_no_submission': Value('bool'), 'seconds': Value('float64')}, 'eligible_error': Value('bool'), 'model': Value('string'), 'split': Value('string')}
because column names don't match
The above exception was the direct cause of the following exception:
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 1880, in _prepare_split_single
raise DatasetGenerationError("An error occurred while generating the dataset") from e
datasets.exceptions.DatasetGenerationError: An error occurred while generating the datasetNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
competition string | trajectory string | stage string | state_id string | parent_state string | task_prompt string | raw_response string | code string | static_task_manifest string | code_sha string | execution dict | eligible_error bool | model string | split string |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
3d-object-detection-for-autonomous-vehicles | 3d-object-detection-for-autonomous-vehicles/init-000 | initial | b3e91197f050 | null | You are solving the Kaggle competition `3d-object-detection-for-autonomous-vehicles`.
# Overview
## Description
Self-driving technology presents a rare opportunity to improve the quality of life in many o... | ```python
import pandas as pd
import numpy as np
import json
import os
from pathlib import Path
from PIL import Image
import torch
import torch.nn as nn
from torch.utils.data import Dataset, DataLoader
from torchvision import models
import warnings
warnings.filterwarnings('ignore')
# Set random seeds for reproducibili... | import pandas as pd
import numpy as np
import json
import os
from pathlib import Path
from PIL import Image
import torch
import torch.nn as nn
from torch.utils.data import Dataset, DataLoader
from torchvision import models
import warnings
warnings.filterwarnings('ignore')
# Set random seeds for reproducibility
np.rand... | Authoritative static facts from the fixed task environment, collected before candidate execution:
directory entries=description.md,sample_submission.csv,test_data,test_images,test_lidar,test_maps,train.csv,train_data,train_images,train_lidar,train_maps
sample_submission.csv: CSV; columns=Id,PredictionString; dtypes str... | 1a4b7e8634dc | {
"returncode": 0,
"timed_out": false,
"exception_class": null,
"exception_message": null,
"stdout": "Training model...\nEpoch 1, Loss: -2.4150\nEpoch 2, Loss: -6.5038\nGenerating predictions...\nSubmission saved to /scratch/work/pr/3d-object-detection-for-autonomous-vehicles/init-000/submission.csv\nSubmissi... | false | claude-haiku-4-5 | eval |
3d-object-detection-for-autonomous-vehicles | 3d-object-detection-for-autonomous-vehicles/init-001 | initial | 74cd5f918e00 | null | You are solving the Kaggle competition `3d-object-detection-for-autonomous-vehicles`.
# Overview
## Description
Self-driving technology presents a rare opportunity to improve the quality of life in many o... | ```python
import os
import json
import pandas as pd
import numpy as np
from pathlib import Path
import torch
import torch.nn as nn
from torch.utils.data import Dataset, DataLoader
from PIL import Image
from sklearn.preprocessing import StandardScaler
import warnings
warnings.filterwarnings('ignore')
# Set up paths
DAT... | import os
import json
import pandas as pd
import numpy as np
from pathlib import Path
import torch
import torch.nn as nn
from torch.utils.data import Dataset, DataLoader
from PIL import Image
from sklearn.preprocessing import StandardScaler
import warnings
warnings.filterwarnings('ignore')
# Set up paths
DATA_DIR = Pa... | Authoritative static facts from the fixed task environment, collected before candidate execution:
directory entries=description.md,sample_submission.csv,test_data,test_images,test_lidar,test_maps,train.csv,train_data,train_images,train_lidar,train_maps
sample_submission.csv: CSV; columns=Id,PredictionString; dtypes str... | fbf2dd4537f4 | {
"returncode": 0,
"timed_out": false,
"exception_class": null,
"exception_message": null,
"stdout": "Submission saved to /scratch/work/pr/3d-object-detection-for-autonomous-vehicles/init-001/submission.csv\nSubmission shape: (4536, 2)\nSample rows:\n Id Predic... | false | claude-haiku-4-5 | eval |
3d-object-detection-for-autonomous-vehicles | 3d-object-detection-for-autonomous-vehicles/init-002 | initial | 74e522b9a8c2 | null | You are solving the Kaggle competition `3d-object-detection-for-autonomous-vehicles`.
# Overview
## Description
Self-driving technology presents a rare opportunity to improve the quality of life in many o... | ```python
import os
import json
import csv
import numpy as np
import pandas as pd
from pathlib import Path
from collections import defaultdict
import torch
import torch.nn as nn
from torch.utils.data import Dataset, DataLoader
from torchvision import transforms
from PIL import Image
import warnings
warnings.filterwarni... | import os
import json
import csv
import numpy as np
import pandas as pd
from pathlib import Path
from collections import defaultdict
import torch
import torch.nn as nn
from torch.utils.data import Dataset, DataLoader
from torchvision import transforms
from PIL import Image
import warnings
warnings.filterwarnings('ignor... | Authoritative static facts from the fixed task environment, collected before candidate execution:
directory entries=description.md,sample_submission.csv,test_data,test_images,test_lidar,test_maps,train.csv,train_data,train_images,train_lidar,train_maps
sample_submission.csv: CSV; columns=Id,PredictionString; dtypes str... | 205b423853ed | {
"returncode": 0,
"timed_out": false,
"exception_class": null,
"exception_message": null,
"stdout": "Classes found: []\nSubmission written to /scratch/work/pr/3d-object-detection-for-autonomous-vehicles/init-002/submission.csv\nTotal predictions: 4536\n Id ... | false | claude-haiku-4-5 | eval |
3d-object-detection-for-autonomous-vehicles | 3d-object-detection-for-autonomous-vehicles/init-003 | initial | 977c3356c8c6 | null | You are solving the Kaggle competition `3d-object-detection-for-autonomous-vehicles`.
# Overview
## Description
Self-driving technology presents a rare opportunity to improve the quality of life in many o... | ```python
import os
import json
import csv
import numpy as np
import pandas as pd
from pathlib import Path
from PIL import Image
import torch
import torch.nn as nn
from torch.utils.data import Dataset, DataLoader
from torchvision import transforms
from sklearn.preprocessing import StandardScaler
import warnings
warning... | import os
import json
import csv
import numpy as np
import pandas as pd
from pathlib import Path
from PIL import Image
import torch
import torch.nn as nn
from torch.utils.data import Dataset, DataLoader
from torchvision import transforms
from sklearn.preprocessing import StandardScaler
import warnings
warnings.filterwa... | Authoritative static facts from the fixed task environment, collected before candidate execution:
directory entries=description.md,sample_submission.csv,test_data,test_images,test_lidar,test_maps,train.csv,train_data,train_images,train_lidar,train_maps
sample_submission.csv: CSV; columns=Id,PredictionString; dtypes str... | f1c1c8032fd8 | {
"returncode": 0,
"timed_out": false,
"exception_class": null,
"exception_message": null,
"stdout": "Submission saved to /scratch/work/pr/3d-object-detection-for-autonomous-vehicles/init-003/submission.csv\nTotal samples: 37296\nSamples with predictions: 0\n",
"stderr": "",
"submission_exists": true,
"... | false | claude-haiku-4-5 | eval |
alaska2-image-steganalysis | alaska2-image-steganalysis/init-000 | initial | d0e3c7d5a96a | null | You are solving the Kaggle competition `alaska2-image-steganalysis`.
# Overview
## Description
That file you downloaded may contain hidden messages that aren't part of its regular contents. The same technology employed for digital watermarking is also misused by crime rings. Law enforcement must now use steganalysis... | ```python
import os
import numpy as np
import pandas as pd
from PIL import Image
import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import Dataset, DataLoader
import torchvision.transforms as transforms
from torchvision.models import resnet50
import warnings
warnings.filterwarnings('ig... | import os
import numpy as np
import pandas as pd
from PIL import Image
import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import Dataset, DataLoader
import torchvision.transforms as transforms
from torchvision.models import resnet50
import warnings
warnings.filterwarnings('ignore')
# ... | Authoritative static facts from the fixed task environment, collected before candidate execution:
directory entries=Cover,JMiPOD,JUNIWARD,Test,UERD,description.md,sample_submission.csv
sample_submission.csv: CSV; columns=Id,Label; dtypes string=Id; int64=Label; no nulls in 20k-row sample | b0289589a07b | {
"returncode": 1,
"timed_out": false,
"exception_class": "RuntimeError",
"exception_message": "DataLoader worker (pid(s) 265) exited unexpectedly",
"stdout": "Loading training data...\nTraining samples: 9998\nTraining model...\n",
"stderr": "ERROR: Unexpected bus error encountered in worker. This might be ... | false | claude-haiku-4-5 | eval |
alaska2-image-steganalysis | alaska2-image-steganalysis/init-000 | repair-1 | 2decb443bca1 | d0e3c7d5a96a | You are solving the Kaggle competition `alaska2-image-steganalysis`.
# Overview
## Description
That file you downloaded may contain hidden messages that aren't part of its regular contents. The same technology employed for digital watermarking is also misused by crime rings. Law enforcement must now use steganalysis... | ```python
import os
import numpy as np
import pandas as pd
from PIL import Image
import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import Dataset, DataLoader
import torchvision.transforms as transforms
from torchvision.models import resnet50
import warnings
warnings.filterwarnings('ig... | import os
import numpy as np
import pandas as pd
from PIL import Image
import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import Dataset, DataLoader
import torchvision.transforms as transforms
from torchvision.models import resnet50
import warnings
warnings.filterwarnings('ignore')
# ... | Authoritative static facts from the fixed task environment, collected before candidate execution:
directory entries=Cover,JMiPOD,JUNIWARD,Test,UERD,description.md,sample_submission.csv
sample_submission.csv: CSV; columns=Id,Label; dtypes string=Id; int64=Label; no nulls in 20k-row sample | 006f74505507 | {
"returncode": 0,
"timed_out": false,
"exception_class": null,
"exception_message": null,
"stdout": "Loading training data...\nTraining samples: 6000\nTraining model...\nEpoch 1, Batch 30, Loss: 0.6919\nEpoch 1, Batch 60, Loss: 0.6449\nEpoch 1, Batch 90, Loss: 0.7070\nEpoch 1, Batch 120, Loss: 0.7864\nEpoch ... | false | claude-haiku-4-5 | eval |
alaska2-image-steganalysis | alaska2-image-steganalysis/init-001 | initial | b10d06ae21d0 | null | You are solving the Kaggle competition `alaska2-image-steganalysis`.
# Overview
## Description
That file you downloaded may contain hidden messages that aren't part of its regular contents. The same technology employed for digital watermarking is also misused by crime rings. Law enforcement must now use steganalysis... | ```python
import os
import numpy as np
import pandas as pd
from pathlib import Path
from PIL import Image
import torch
import torch.nn as nn
import torchvision.transforms as transforms
from torchvision import models
from torch.utils.data import Dataset, DataLoader
import torch.optim as optim
from sklearn.metrics import... | import os
import numpy as np
import pandas as pd
from pathlib import Path
from PIL import Image
import torch
import torch.nn as nn
import torchvision.transforms as transforms
from torchvision import models
from torch.utils.data import Dataset, DataLoader
import torch.optim as optim
from sklearn.metrics import roc_auc_s... | Authoritative static facts from the fixed task environment, collected before candidate execution:
directory entries=Cover,JMiPOD,JUNIWARD,Test,UERD,description.md,sample_submission.csv
sample_submission.csv: CSV; columns=Id,Label; dtypes string=Id; int64=Label; no nulls in 20k-row sample | 57a3e29197ef | {
"returncode": 0,
"timed_out": false,
"exception_class": null,
"exception_message": null,
"stdout": "Using device: cuda\nLoading training data...\nTraining set size: 5000\nStego count: 3753, Cover count: 1247\nSetting up model...\nTraining...\nEpoch 1/3, Loss: 0.5880\nEpoch 2/3, Loss: 0.5571\nEpoch 3/3, Loss... | false | claude-haiku-4-5 | eval |
alaska2-image-steganalysis | alaska2-image-steganalysis/init-002 | initial | 43f4d5401ef1 | null | "You are solving the Kaggle competition `alaska2-image-steganalysis`.\n\n# Overview\n\n## Descriptio(...TRUNCATED) | "```python\nimport os\nimport torch\nimport torch.nn as nn\nimport torch.optim as optim\nfrom torch.(...TRUNCATED) | "import os\nimport torch\nimport torch.nn as nn\nimport torch.optim as optim\nfrom torch.utils.data (...TRUNCATED) | "Authoritative static facts from the fixed task environment, collected before candidate execution:\n(...TRUNCATED) | 920411851c61 | {"returncode":1,"timed_out":false,"exception_class":"RuntimeError","exception_message":"DataLoader w(...TRUNCATED) | false | claude-haiku-4-5 | eval |
alaska2-image-steganalysis | alaska2-image-steganalysis/init-002 | repair-1 | cebdc48b5022 | 43f4d5401ef1 | "You are solving the Kaggle competition `alaska2-image-steganalysis`.\n\n# Overview\n\n## Descriptio(...TRUNCATED) | "```python\nimport os\nimport torch\nimport torch.nn as nn\nimport torch.optim as optim\nfrom torch.(...TRUNCATED) | "import os\nimport torch\nimport torch.nn as nn\nimport torch.optim as optim\nfrom torch.utils.data (...TRUNCATED) | "Authoritative static facts from the fixed task environment, collected before candidate execution:\n(...TRUNCATED) | 1a72ed7d0949 | {"returncode":0,"timed_out":false,"exception_class":null,"exception_message":null,"stdout":"Using de(...TRUNCATED) | false | claude-haiku-4-5 | eval |
alaska2-image-steganalysis | alaska2-image-steganalysis/init-003 | initial | e5bb6a4dc69f | null | "You are solving the Kaggle competition `alaska2-image-steganalysis`.\n\n# Overview\n\n## Descriptio(...TRUNCATED) | "```python\nimport os\nimport numpy as np\nimport pandas as pd\nfrom pathlib import Path\nimport tor(...TRUNCATED) | "import os\nimport numpy as np\nimport pandas as pd\nfrom pathlib import Path\nimport torch\nimport (...TRUNCATED) | "Authoritative static facts from the fixed task environment, collected before candidate execution:\n(...TRUNCATED) | 34a779f1b674 | {"returncode":1,"timed_out":false,"exception_class":"ConnectionResetError","exception_message":"[Err(...TRUNCATED) | false | claude-haiku-4-5 | eval |
End of preview.
Agent-CWM fresh rubrics v1
Verifiable code-quality rubrics for ML-engineering programs, mined from agent traces on MLE-bench competitions and judged statically (no execution at judge time). Two families:
- Error rubrics (381): each predicts a runtime failure and names the exception class. Truth = the interpreter's actual exception (subclass-aware).
- Performance rubrics (327): each names a concrete performance-relevant design issue. Truth = adjudicated annotator credits against measured score gaps.
Headline results (frozen split: 24 mine / 37 eval competitions)
| Stage (mine/eval) | Err rubrics | Perf rubrics | Error P | Error catch rate | Perf P | Perf R |
|---|---|---|---|---|---|---|
| 5 / 8 | 74 | 77 | 0.000 | 0.000 | 0.234 | 0.160 |
| 10 / 15 | 137 | 183 | 0.000 | 0.000 | 0.237 | 0.319 |
| 15 / 23 | 188 | 259 | 0.152 | 0.070 | 0.293 | 0.454 |
| 20 / 31 | 278 | 319 | 0.303 | 0.240 | 0.276 | 0.427 |
| 24 / 37 (full) | 381 | 327 | 0.384 [0.31–0.46] | 0.483 | 0.259 [0.19–0.33] | 0.406 |
Error catch rate = fraction of catchable held-out failures (149) where at least one fired rubric named the true exception class. Timeouts (73) and environment kills (7) are excluded by construction — no static rubric can name them. Precision is decision-level with trajectory-bootstrap 95% CIs. Dashboard: https://codingwm.github.io/fresh_pr.html
Layout
rubrics/error/ 381 error rubric .md files (frontmatter: id, exception class, criteria)
rubrics/perf/ 327 performance rubric .md files (protocol style)
rubrics/candidates_perf.jsonl stage-2 perf candidates with provenance
traces/traces_banked.jsonl the full fresh-era trace bank (1,372 states over the
24 mine comps): initial/repair error states AND the
323 perf climb states with graded scores. Every
(state_id, better_state_id) in the perf candidates'
_provenance resolves here — score_worse/score_better
pairs are reproducible from these rows.
traces/err27_<comp>.jsonl mined program states: task_prompt, code, execution
(exception class/message), trajectory structure
curation_logs/ error curation logs (state -> rubric render, truth
exception, emitted/lint/dup decision) — full provenance
eval/eval_states_v5.jsonl held-out eval corpus (37 comps): programs + executions
eval/truth_annotations_final.jsonl perf truth annotations (adjudicated credits)
eval/pr_decisions_final.jsonl every judged (rubric, state) decision cell
results/ final stage table + P/R with CIs + the frozen split
Protocol notes
- Mining sources and eval competitions are disjoint (frozen split in
results/split_used.json); scaling stages are nested prefixes of one split. - Judge: blind static manifest preset (judge never sees execution results).
- Collection: haiku coder, ≤3 repair / ≤8 climb turns, ≤40 programs/comp, stop at 15 curated rubrics per family or stagnation.
- Rubric generation: claude-opus renders, v12 curation prompt (error side).
Part of the agent-cwm / CodeWorld GYM project (rubric-grounded code world models). Contact: Edward Sun (edwardsun12895@g.ucla.edu).
- Downloads last month
- 111