Toward a Compositional Knowledge Architecture for a Shared Organizational Cortex
Paper • 2611.05753 • Published
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/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/harbor/harbor.py", line 171, in _split_generators
raise DataFilesNotFoundError("No task.toml or instruction.md files found")
datasets.exceptions.DataFilesNotFoundError: No task.toml or instruction.md files found
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.
Geometric Knowledge Evaluations
Registered plan, pilot judgments, and retrieval results for Toward a Compositional Knowledge Architecture for a Shared Organizational Cortex (Miracle, Georgiou, Okoth, and Monday, 2026), arXiv:2611.05753.
This dataset is the public evaluation record at commit e699e73 of withrobinhq/geometric-knowledge-paper-evals. The registered plan is tag registered-plan, commit 6aed8eeb9baa03f0f878e111f2f7e6943c51652f, frozen 2026-10-20T16:20:07Z.
Nineteen components are registered. The Dimension term, standing penalty, Domain scoping, and HyDE escalation are scored on 272 pooled held-out queries in arms (i)–(viii). Arm (ix), facet-only, is scored in the same run. The pilot is annotators A and B on two public domains, plus a calibration set. Component status is in execution_record.yaml.
Contents
| Path | What it holds |
|---|---|
evaluation-plan.md |
The evaluation specification. Tag registered-plan (2026-10-20) holds the registered text, SHA-256 413b34d57539da0a8971a86c18a2abf533d9493d4dddf997c60e78e5358278ab. |
AMENDMENTS.md |
Earlier versions of the plan. |
execution_record.yaml |
Status of the nineteen components, the reported contrasts, and arm (ix). |
pilot/ |
A/B judgments (calibration.csv, peps.csv, enron.csv), protocol, provenance, normalized inputs, and pilot_results.csv (149 metric rows). |
pilot/scripts/ |
Scoring and the checks below. |
results/ |
Pooled arm means, gains, ablation arms, and the descriptive pilot validity table. |
The release contains the summary record, the pilot judgments, the normalized inputs, and the scripts.
Use
Requirements: Python 3 with NumPy 2.5.4 and krippendorff 0.9.0.
python pilot/scripts/score.py --selftest
python pilot/scripts/score.py pilot/peps.csv --compare A:B --dataset peps --csv /tmp/peps.csv
python pilot/scripts/score.py pilot/enron.csv --compare A:B --dataset enron --csv /tmp/enron.csv
python pilot/scripts/score.py pilot/calibration.csv --compare A:B --dataset calibration_ab --csv /tmp/calibration_ab.csv
python pilot/scripts/score.py pilot/calibration.csv --compare B:KEY --dataset calibration_blind_key --csv /tmp/calibration_blind_key.csv
python pilot/scripts/verify_results.py --expected pilot/pilot_results.csv /tmp/calibration_ab.csv /tmp/calibration_blind_key.csv /tmp/peps.csv /tmp/enron.csv
python pilot/scripts/verify_release.py
verify_results.py checks the four scored files against pilot/pilot_results.csv. verify_release.py checks results/ and execution_record.yaml against the figures reported in the paper. Retrieval rows are per arm and per contrast. There is no single summary score.
Licences
This repository states no licence. The scorer imports krippendorff 0.9.0 (GPL-3.0-or-later). Selected PEP material is public domain or CC0-1.0-Universal; PEP 600 follows the PEP 1 policy. The CMU Enron release carries no written licence in the source archive. Cite the paper with CITATION.cff. Works cited by the plan are listed at the end of evaluation-plan.md.
Citation
@misc{miracle2026compositionalknowledgearchitectureshared,
title={Toward a Compositional Knowledge Architecture for a Shared Organizational Cortex},
author={Andrew Miracle and Phyl Georgiou and Christine Okoth and Ese Monday},
year={2026},
eprint={2611.05753},
archivePrefix={arXiv},
primaryClass={cs.AI},
url={https://arxiv.org/abs/2611.05753},
}
- Downloads last month
- 67