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/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/webdataset/webdataset.py", line 84, in _split_generators
raise ValueError(
...<2 lines>...
)
ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
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.
MUSTER evaluation artifacts
Saved evaluation outputs for MUSTER: Top-Ranked, Self-Configuring Agent Memory That Knows a Log from a Novel and Retrieves Each by Its Structure.
Code: https://github.com/saintmachines/MUSTER · Project: https://saintmachines.github.io/MUSTER/ Browser structure inspector: https://saintmachines.github.io/MUSTER/try/
The public arXiv identifier will be linked when announced. This repository is a download and provenance package, rather than a homogeneous training dataset; extract the archive to use the supplied verifier and code's offline scorer.
Included
| Artifact | Files | Rows | Original bytes |
|---|---|---|---|
| BEAM reader answers | 51 | 22,100 | 108,294,974 |
| BEAM judgments | 51 | 22,100 | 38,900,825 |
| Honcho query/context artifacts | 6 | 2,600 | 216,270,040 |
The two Honcho formats describe the same 1,300 context requests. BEAM runs are distinct run tags; the row totals are not a count of unique benchmark questions. No new model responses or judgments were generated for this release.
MUSTER_eval_v1.tar.gz: 77,486,652 bytes, SHA-256 617c19d23c02a8bc958fbc5dba7d6e41fae11b8c672a5a96df5b562200bc77f7. It is the same file as the asset of the GitHub release paper-artifacts-v1. Coverage and offline verification describe this bundle.
Use
Verify the download against SHA256SUMS, extract it, and run:
python MUSTER_eval_v1/verify_eval_artifacts.py MUSTER_eval_v1
From the MUSTER code repository:
MUSTER_HOME=/absolute/path/to/MUSTER_eval_v1 python gate/rescore_beam.py
All 48 printed values of the main BEAM table reproduce from 51 saved judgment files, without a model/API call. This is re-scoring saved judgments, not regenerating answers or judgments. The archive includes the verification report, manifest, scorer output and license texts.
Provenance and scope
Every result file retains its original bytes. Answer/judgment IDs were joined by (row, qi). All 1,300 saved Honcho contexts were checked by ID against the original frozen questions and companion rendered contexts. The JSONL contains saved API context messages; the JSON joins those messages to the source questions. These are query/context artifacts, not raw HTTP captures, intermediate retries or historical ingestion responses. The public harness documents the context call and ingestion configuration.
MUSTER's original annotations and verifier code use Apache-2.0. BEAM source content retains CC BY-SA 4.0 as declared by the BEAM and BEAM-10M dataset cards. LICENSE_BEAM supplies attribution and license links and preserves the original BEAM software MIT notice separately. BEAM source: https://github.com/mohammadtavakoli78/BEAM . Other benchmarks and model weights are outside this archive. The synthetic/code conversations can include example email/token-like strings from the original benchmark.
Limit: a hosted service is not fully version-pinned by retaining contexts; this archive preserves the observed retrieval outputs used in the evaluation. See the paper for prompts, budgets, aggregation and limitations.
- Downloads last month
- 16