Dataset Viewer
The dataset viewer is not available for this subset.
Cannot get the split names for the config 'default' of the dataset.
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 291, 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 306, 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.
CodeNib rankings on Agent Retrieval Bench
Full file rankings from eight CodeNib retrieval runs on all 345 positive samples of Agent Retrieval Bench (ARB). The files exist so anyone can rescore the results with ARB's own evaluator without installing CodeNib.
| File | Contents |
|---|---|
rankings/<run>_predictions.json.gz |
One record per sample: sample_id, repo, base_commit, status, the complete ranked_files, and per-sample cost fields |
rankings/SHA256SUMS |
Checksums of the ranking files |
arb_full.json |
Run configurations, per-sample metrics in metric_order, ranking depths, cost summaries, and the runs' implementation receipt |
rescore.py |
Recomputes every metric from the rankings with ARB's evaluator and compares it with arb_full.json |
Runs
| Run | Method | Source |
|---|---|---|
bm25 |
CodeNib BM25 over every released file and symbol chunk | Released corpus |
zoekt |
Zoekt trigram queries for the code-like terms in the query | Released corpus |
dense |
Qwen3-Embedding-0.6B with its query instruction and a 2,048-token limit | Released corpus |
dense_4b |
Qwen3-Embedding-4B with the same settings | Released corpus |
hybrid |
File-level RRF of the bm25 and dense rankings |
Released corpus |
rerank |
The top 50 dense files rescored by Qwen3-Reranker-4B and fused with the dense order by RRF |
Released corpus |
rerank_4b |
The same over dense_4b |
Released corpus |
graph_expansion |
The top BM25 files, then files linked to them in a code graph | Full source |
graph_expansion also indexes a depth-one checkout of each sample's
base_commit, which is ARB's separately labeled full-source condition; its
rankings may name files outside the released corpus, and rescore.py drops
them before cutoffs. The methods and the full-source rules are described in
CodeNib's ARB guide
and ARB's reporting contract.
Versions
| Item | Value |
|---|---|
| CodeNib | tag arb/2026-10-24/methods |
| ARB scoring code | 07014c986f3deadb1548c62b32c0ffbe6a81465d |
| Dataset | eyuansu71/agent_retrieval_bench at 5901e1ee3aff048290db72edf9c63bc498b79ea3, releases v2_code2test, v2_comment2context, v2_trace2code, v2_edit2ripple |
| Embedding models | Qwen/Qwen3-Embedding-0.6B at 97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3; Qwen/Qwen3-Embedding-4B at 5cf2132abc99cad020ac570b19d031efec650f2b |
| Reranker | Qwen/Qwen3-Reranker-4B at 22e683669bc0f0bd69640a1354a6d0aebcfeede5 |
The earlier five-method rankings, from tag arb/2026-10-22/methods, remain at
revision 8be160b7.
Rescore
From the root of an ARB checkout, at 07014c98 or any later revision whose
scoring code is unchanged:
for release in v2_code2test v2_comment2context v2_trace2code v2_edit2ripple; do
arb download-benchmark --version "$release" \
--revision 5901e1ee3aff048290db72edf9c63bc498b79ea3 --local-dir data
done
hf download sysevol-ai/codenib-arb-rankings --repo-type dataset --local-dir codenib-arb
(cd codenib-arb/rankings && sha256sum -c SHA256SUMS)
python codenib-arb/rescore.py --rankings codenib-arb/rankings --expected codenib-arb/arb_full.json
It needs no model, no API key, and no network after the download, and prints the per-run means and how many per-sample metric cells match.
- Downloads last month
- 87