Datasets:

moorcheh
/

mair-moorcheh-new-vs-old-latency

Languages:

English

ArXiv:

License:

Dataset card Data Studio Files Files and versions

xet

Community

Dataset Preview

Duplicate

Split (9)

comparison_summary

The full dataset viewer is not available (click to read why). Only showing a preview of the rows.

The dataset generation failed because of a cast error

Error code:   DatasetGenerationCastError
Exception:    DatasetGenerationCastError
Message:      An error occurred while generating the dataset

All the data files must have the same columns, but at some point there are 13 new columns ({'Moorcheh_validate_namespace_mean_ms', 'Moorcheh_fetch_complete_data_mean_ms', 'Moorcheh_select_candidates_mean_ms', 'Moorcheh_parse_validate_mean_ms', 'Search_Server_Mean_ms', 'Moorcheh_fetch_data_mean_ms', 'Moorcheh_apply_metadata_filter_mean_ms', 'Moorcheh_reorder_filter_mean_ms', 'Moorcheh_format_response_mean_ms', 'Moorcheh_authorize_mean_ms', 'Moorcheh_calculate_distance_mean_ms', 'Moorcheh_prepare_vector_mean_ms', 'Moorcheh_calculate_scores_mean_ms'}) and 15 missing columns ({'Change %', 'Old Median ms', 'New Mean ms', 'New Median ms', 'Moorcheh_calculate_distance_mean_ms New', 'Moorcheh_fetch_data_mean_ms New', 'Moorcheh_calculateDistance_mean_ms Old', 'Moorcheh_authorize_mean_ms New', 'Change Type', 'Category', 'Search_Server_Mean_ms New', 'Moorcheh_authorize_mean_ms Old', 'Old Mean ms', 'Moorcheh_fetchData_mean_ms Old', 'Search_Server_Mean_ms Old'}).

This happened while the csv dataset builder was generating data using

hf://datasets/moorcheh/mair-moorcheh-new-vs-old-latency/k10_vs_k100.csv (at revision 17f97fcc7cb42d3088340cde328b3491cff96fe2)

Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1831, in _prepare_split_single
                  writer.write_table(table)
                File "/usr/local/lib/python3.12/site-packages/datasets/arrow_writer.py", line 714, in write_table
                  pa_table = table_cast(pa_table, self._schema)
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2272, in table_cast
                  return cast_table_to_schema(table, schema)
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2218, in cast_table_to_schema
                  raise CastError(
              datasets.table.CastError: Couldn't cast
              Dataset: string
              Num_Corpus: int64
              Search_Server_Mean_ms: double
              Moorcheh_authorize_mean_ms: double
              Moorcheh_parse_validate_mean_ms: double
              Moorcheh_validate_namespace_mean_ms: double
              Moorcheh_prepare_vector_mean_ms: double
              Moorcheh_fetch_data_mean_ms: double
              Moorcheh_calculate_distance_mean_ms: double
              Moorcheh_select_candidates_mean_ms: double
              Moorcheh_calculate_scores_mean_ms: double
              Moorcheh_fetch_complete_data_mean_ms: double
              Moorcheh_apply_metadata_filter_mean_ms: double
              Moorcheh_reorder_filter_mean_ms: double
              Moorcheh_format_response_mean_ms: double
              -- schema metadata --
              pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 2624
              to
              {'Category': Value('string'), 'Old Mean ms': Value('float64'), 'New Mean ms': Value('float64'), 'Change %': Value('float64'), 'Change Type': Value('string'), 'Old Median ms': Value('float64'), 'New Median ms': Value('float64'), 'Dataset': Value('string'), 'Num_Corpus': Value('int64'), 'Moorcheh_authorize_mean_ms New': Value('float64'), 'Moorcheh_authorize_mean_ms Old': Value('float64'), 'Moorcheh_fetch_data_mean_ms New': Value('float64'), 'Moorcheh_fetchData_mean_ms Old': Value('float64'), 'Moorcheh_calculate_distance_mean_ms New': Value('float64'), 'Moorcheh_calculateDistance_mean_ms Old': Value('float64'), 'Search_Server_Mean_ms New': Value('float64'), 'Search_Server_Mean_ms Old': Value('float64')}
              because column names don't match
              
              During handling of the above exception, another exception occurred:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1339, in compute_config_parquet_and_info_response
                  parquet_operations = convert_to_parquet(builder)
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 972, in convert_to_parquet
                  builder.download_and_prepare(
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 894, in download_and_prepare
                  self._download_and_prepare(
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 970, in _download_and_prepare
                  self._prepare_split(split_generator, **prepare_split_kwargs)
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1702, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1833, in _prepare_split_single
                  raise DatasetGenerationCastError.from_cast_error(
              datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
              
              All the data files must have the same columns, but at some point there are 13 new columns ({'Moorcheh_validate_namespace_mean_ms', 'Moorcheh_fetch_complete_data_mean_ms', 'Moorcheh_select_candidates_mean_ms', 'Moorcheh_parse_validate_mean_ms', 'Search_Server_Mean_ms', 'Moorcheh_fetch_data_mean_ms', 'Moorcheh_apply_metadata_filter_mean_ms', 'Moorcheh_reorder_filter_mean_ms', 'Moorcheh_format_response_mean_ms', 'Moorcheh_authorize_mean_ms', 'Moorcheh_calculate_distance_mean_ms', 'Moorcheh_prepare_vector_mean_ms', 'Moorcheh_calculate_scores_mean_ms'}) and 15 missing columns ({'Change %', 'Old Median ms', 'New Mean ms', 'New Median ms', 'Moorcheh_calculate_distance_mean_ms New', 'Moorcheh_fetch_data_mean_ms New', 'Moorcheh_calculateDistance_mean_ms Old', 'Moorcheh_authorize_mean_ms New', 'Change Type', 'Category', 'Search_Server_Mean_ms New', 'Moorcheh_authorize_mean_ms Old', 'Old Mean ms', 'Moorcheh_fetchData_mean_ms Old', 'Search_Server_Mean_ms Old'}).
              
              This happened while the csv dataset builder was generating data using
              
              hf://datasets/moorcheh/mair-moorcheh-new-vs-old-latency/k10_vs_k100.csv (at revision 17f97fcc7cb42d3088340cde328b3491cff96fe2)
              
              Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)

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.

Category string	Old Mean ms float64	New Mean ms float64	Change % float64	Change Type string	Old Median ms float64	New Median ms float64	Dataset string	Num_Corpus int64	Moorcheh_authorize_mean_ms New float64	Moorcheh_authorize_mean_ms Old float64	Moorcheh_fetch_data_mean_ms New float64	Moorcheh_fetchData_mean_ms Old float64	Moorcheh_calculate_distance_mean_ms New float64	Moorcheh_calculateDistance_mean_ms Old float64	Search_Server_Mean_ms New float64	Search_Server_Mean_ms Old float64
corpusSearch	525.66	210.36	-60	✓ Improvement	242.49	191.77	Apple	678	37.55	46.3	12.52	20.61	0.6	9.78	165.62	113.16
authorize	89.77	42.24	-53	✓ Improvement	50.81	39.29	HC3Finance	3,933	37.05	48.51	18.95	97.38	3.14	34.04	175.88	206.86
parseValidate	0.49	0.55	12.2	✗ Regression	0.49	0.55	ConvFinQA	6,503	37.03	52.95	29.44	167.11	5.3	58.23	192.39	308.85
validateNamespace	4.16	2.88	-30.8	✓ Improvement	4.19	2.8	FinQA	11,865	46	49.97	51.56	321.86	9.5	103.79	234.85	505.87
prepareVector	0.01	0.13	1,200	✗ Regression	0.01	0.13	FinanceBench	15,325	44.13	45.16	65.38	418.01	11.8	206.67	246.16	713.92
fetchData	358.55	51.89	-85.5	✓ Improvement	100.25	28.18	FiQA	57,638	36.6	53.44	209.54	1,533.07	46.93	802.89	462.48	2,453.65
calculateDistance	147.92	9.66	-93.5	✓ Improvement	53.15	5.41	AILA2019-Statutes	197	54.97	51.85	9.15	10.3	0.21	2.85	118.37	83.89
selectCandidates	3.41	7.24	112.3	✗ Regression	1.47	2.69	AILA2019-Case	2,914	52.94	56.98	19.78	89.98	2.52	42.16	190.55	230.34
calculateScores	26.61	13.59	-49	✓ Improvement	24.38	13.27	LeCaRDv2	3,000	53.21	54.7	17.89	78.21	2.53	26.31	197.97	186.28
applyMetadataFilter	0.01	0.04	300	✗ Regression	0.01	0.04	REGIR-UK2EU	3,930	36.06	49.67	19.21	110.51	3.19	54.99	172.92	254.63
reorderFilter	0.51	0.14	-72.5	✓ Improvement	0.56	0.15	REGIR-EU2UK	10,000	36.6	45.51	41.69	269.52	7.92	140.46	205.65	499.41
formatResponse	0.15	0.02	-86.7	✓ Improvement	0.14	0.02	LegalQuAD	17,702	35.73	45.22	72.6	411.61	13.57	151.48	247.47	642.78
null	null	null	null	null	null	null	ACORDAR	31,589	47.98	54.58	131.57	869.32	25.73	269.31	350.63	1,231.8
null	null	null	null	null	null	null	NFCorpus	3,633	35.61	52.46	17.11	92.51	2.87	31.6	170.55	204.47

Moorcheh Architecture Evolution: New vs Old Latency

This repository provides a deep dive into the performance gains of the New Moorcheh Architecture compared to the legacy version using the MAIR (Massive Instructed Retrieval) methodology.

📂 Dataset Splits

You can toggle between the 9 result sets using the Viewer at the top of this page.

Split	Description
`comparison_summary`	Overall speedup percentages.
`mair_new_full`	Complete metrics for the latest version.
`timing_10gb`	Latency results for 10GB dataset scaling.
`k10_vs_k100`	Analysis of depth impact on speed.

🛠 Python Usage

from datasets import load_dataset
# Load the comparison summary
ds = load_dataset("moorcheh/mair-moorcheh-new-vs-old-latency", split="comparison_summary")
print(ds.to_pandas().head())

Citation

If you use this dataset, please cite:

arXiv:2601.11557

Downloads last month: 12

Total file size:

40.9 kB

Paper for moorcheh/mair-moorcheh-new-vs-old-latency

From HNSW to Information-Theoretic Binarization: Rethinking the Architecture of Scalable Vector Search

Paper • 2601.11557 • Published Dec 16, 2025 • 3