The dataset viewer is not available for this subset.
Exception: SplitsNotFoundError
Message: The split names could not be parsed from the dataset config.
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 290, in _generate_tables
pa_table = paj.read_json(
io.BytesIO(batch), read_options=paj.ReadOptions(block_size=block_size)
)
File "pyarrow/_json.pyx", line 364, in pyarrow._json.read_json
File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
raise convert_status(status)
pyarrow.lib.ArrowInvalid: JSON parse error: Column() changed from object to string in row 0
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
for split_generator in builder._split_generators(
~~~~~~~~~~~~~~~~~~~~~~~~~^
StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 101, in _split_generators
pa_table = next(iter(self._generate_tables(**splits[0].gen_kwargs, allow_full_read=False)))[1]
~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 304, in _generate_tables
batch = json_encode_fields_in_json_lines(original_batch, json_field_paths)
File "/usr/local/lib/python3.14/site-packages/datasets/utils/json.py", line 111, in json_encode_fields_in_json_lines
examples = [ujson_loads(line) for line in original_batch.splitlines()]
~~~~~~~~~~~^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/utils/json.py", line 20, in ujson_loads
return pd.io.json.ujson_loads(*args, **kwargs)
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
ValueError: Expected object or value
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/split_names.py", line 68, in compute_split_names_from_streaming_response
for split in get_dataset_split_names(
~~~~~~~~~~~~~~~~~~~~~~~^
path=dataset,
^^^^^^^^^^^^^
config_name=config,
^^^^^^^^^^^^^^^^^^^
token=hf_token,
^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
info = get_dataset_config_info(
path,
...<6 lines>...
**config_kwargs,
)
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Molecule Editing Agents datasets
Frozen inputs and retrieval evidence for the five-method experiments in
charlieqi02/Molecule-Editing-Agents.
This repository contains 42 data/notice files totaling 3,839,879,858 bytes (about
3.58 GiB), plus this card and a SHA-256 manifest. It retains the code repository's
datasets/ layout so downloads need no manual renaming.
Contents
| Category | Included artifacts |
|---|---|
| MOSES | Original dataset_v1.csv; six-property train.csv, test.csv, test_scaffold.csv, preprocessing manifest and error record; the exact frozen 4,000-question test sample. |
| ChemCoTBench | The local 1,495-case benchmark and its upstream README. The experiment launchers use all 600 mol_opt cases: 100 for each of six objectives. |
| ChEMBL retrieval | Three production ChEMBL 37 SQLite indices: molecule features, molecular structures, and functional groups, with upstream attribution/license notices. |
| Skill databases | Pair and plan SQLite banks for no_teacher_hint and teacher_hint; F2R training histories and their directional re-evaluations under skills/f2r/. |
Each SQLite file is a consistent snapshot produced by SQLite's backup API,
including committed WAL data, and passed PRAGMA quick_check. Lock files, WAL/SHM
sidecars, smoke indices, raw ChEMBL archives, unrelated datasets, experiment logs,
code, demo assets, credentials, and property-model pickles are excluded.
Both evaluation models use the same GPT-6-Luna F2R evidence by default: 200 MOSES
training histories and 832 directional property-specific views. The optional
Claude-Haiku evidence contains 200 training histories and 877 directional views.
Directional evidence is a measured re-evaluation of existing training attempts,
not new ChemCoTBench test trajectories. See the bank's embedded provenance and
the code's docs/chemcot_f2r_history.md for the conversion procedure. These files
do not contain new GPT-6-Sol or Claude Opus 5.5 experiment results.
Download into a code checkout
Follow handsoff.md. The dataset repository currently requires authorized Hugging Face access.
bash scripts/setup_environment.bash
.venvs/agentic-benchmark/bin/hf auth login
.venvs/agentic-benchmark/bin/python scripts/download_datasets.py
.venvs/agentic-benchmark/bin/python scripts/prepare_oracles.py
The code pins a dataset commit in configs/dataset_release.json.
download_datasets.py downloads only manifest-listed dataset paths and verifies
all SHA-256 digests. It does not overwrite the code checkout's README.
prepare_oracles.py separately downloads the three pretrained PyTDC classifiers
from their upstream source and checks their hashes against the MOSES manifest.
Sources and licenses
This collection does not replace the source artifacts' individual terms.
- MOSES comes from molecularsets/moses,
data/dataset_v1.csv; the original file SHA-256 isbb47a94d347afd476d3828b5e26dceeabc42a2d8cf92a791d00349f22fea0d8b. See the upstream repository for its license and citation. - ChemCoTBench originates from IDEA-AI4S/ChemCoTBench.
The local files were obtained from the public
little1d/MolActmirror at commit0a9daca34747eab291a20c1e28995d4eb27bdd8c; its retained README declares MIT metadata and contains the publication citation. The source log notes a trailing-newline difference inmol_opt/logp.json. - The retrieval databases derive from ChEMBL 37 and carry its CC BY-SA 3.0
notices in
datasets/derived/agentic/retrieval/chembl_notices/. ChEMBL identifiers and release context are preserved. Required citation: Mendez et al., ChEMBL: towards direct deposition of bioassay data, Nucleic Acids Research 47(D1):D930–D940 (2019), DOI: 10.1093/nar/gky1075. - Skill and history files are project-generated retrieval evidence. Their source tasks, source model names, measured outcomes, and provenance remain in the artifacts; effects describe their recorded source context.
Measurements use RDKit LogP/QED, the repository's ChemCoTBench ESOL implementation, and the original PyTDC DRD2/JNK3/GSK3B classifiers. Numerical target matching on MOSES and directional optimization on ChemCoTBench are distinct task contracts.
- Downloads last month
- 39