Dataset Preview
Duplicate
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 2 new columns ({'family', 'correct'}) and 5 missing columns ({'mean_latency_ms', 'preferred_match', 'route_correct', 'errored', 'mean_resp_chars'}).

This happened while the csv dataset builder was generating data using

hf://datasets/moebiusT7/gemma4-quant-regime-study/data/family_rates.csv (at revision 21bbd758241a027b9153f1b27f924738e0504322), ['hf://datasets/moebiusT7/gemma4-quant-regime-study@21bbd758241a027b9153f1b27f924738e0504322/data/cell_metrics.csv', 'hf://datasets/moebiusT7/gemma4-quant-regime-study@21bbd758241a027b9153f1b27f924738e0504322/data/family_rates.csv', 'hf://datasets/moebiusT7/gemma4-quant-regime-study@21bbd758241a027b9153f1b27f924738e0504322/data/paired_quant_effect.csv', 'hf://datasets/moebiusT7/gemma4-quant-regime-study@21bbd758241a027b9153f1b27f924738e0504322/data/regime_output_divergence.csv']

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/python3.14/site-packages/datasets/builder.py", line 1837, in _prepare_split_single
                  writer.write_table(table)
                  ~~~~~~~~~~~~~~~~~~^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 765, in write_table
                  self._write_table(pa_table, writer_batch_size=writer_batch_size)
                  ~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 773, in _write_table
                  pa_table = table_cast(pa_table, self._schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2369, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2297, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              size: string
              regime: string
              suite: string
              family: string
              n: int64
              correct: int64
              rate: double
              -- schema metadata --
              pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 1039
              to
              {'size': Value('string'), 'regime': Value('string'), 'suite': Value('string'), 'n': Value('int64'), 'errored': Value('int64'), 'route_correct': Value('int64'), 'rate': Value('float64'), 'preferred_match': Value('float64'), 'mean_latency_ms': Value('float64'), 'mean_resp_chars': 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 1369, in compute_config_parquet_and_info_response
                  parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
                                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      builder, max_dataset_size_bytes=max_dataset_size_bytes
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
                  builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
                  ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1683, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~^
                      gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  ):
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1839, in _prepare_split_single
                  raise DatasetGenerationCastError.from_cast_error(
                  ...<4 lines>...
                  )
              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 2 new columns ({'family', 'correct'}) and 5 missing columns ({'mean_latency_ms', 'preferred_match', 'route_correct', 'errored', 'mean_resp_chars'}).
              
              This happened while the csv dataset builder was generating data using
              
              hf://datasets/moebiusT7/gemma4-quant-regime-study/data/family_rates.csv (at revision 21bbd758241a027b9153f1b27f924738e0504322), ['hf://datasets/moebiusT7/gemma4-quant-regime-study@21bbd758241a027b9153f1b27f924738e0504322/data/cell_metrics.csv', 'hf://datasets/moebiusT7/gemma4-quant-regime-study@21bbd758241a027b9153f1b27f924738e0504322/data/family_rates.csv', 'hf://datasets/moebiusT7/gemma4-quant-regime-study@21bbd758241a027b9153f1b27f924738e0504322/data/paired_quant_effect.csv', 'hf://datasets/moebiusT7/gemma4-quant-regime-study@21bbd758241a027b9153f1b27f924738e0504322/data/regime_output_divergence.csv']
              
              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.

size
string
regime
string
suite
string
n
int64
errored
int64
route_correct
int64
rate
float64
preferred_match
float64
mean_latency_ms
float64
mean_resp_chars
float64
12B
PTQ
smoke
100
0
88
0.88
0.57
2,954.5
790.9
12B
PTQ
core500
500
0
380
0.76
0.544
2,481.5
750.6
12B
QAT
smoke
100
0
86
0.86
0.58
2,888
857.8
12B
QAT
core500
500
0
374
0.748
0.542
2,626.3
834.3
26B
PTQ
smoke
100
0
83
0.83
0.58
3,739.4
922.4
26B
PTQ
core500
500
0
369
0.738
0.54
3,251.8
874.8
26B
QAT
smoke
100
0
86
0.86
0.56
1,862.5
926.1
26B
QAT
core500
500
0
378
0.756
0.54
1,613.4
898.6
12B
PTQ
smoke
5
null
null
0.8
null
null
null
12B
PTQ
smoke
10
null
null
0.9
null
null
null
12B
PTQ
smoke
10
null
null
1
null
null
null
12B
PTQ
smoke
25
null
null
1
null
null
null
12B
PTQ
smoke
15
null
null
0.9333
null
null
null
12B
PTQ
smoke
35
null
null
0.7429
null
null
null
12B
PTQ
core500
25
null
null
0.56
null
null
null
12B
PTQ
core500
50
null
null
0.66
null
null
null
12B
PTQ
core500
50
null
null
0.92
null
null
null
12B
PTQ
core500
125
null
null
0.984
null
null
null
12B
PTQ
core500
75
null
null
0.9467
null
null
null
12B
PTQ
core500
175
null
null
0.5314
null
null
null
12B
QAT
smoke
5
null
null
0.8
null
null
null
12B
QAT
smoke
10
null
null
0.9
null
null
null
12B
QAT
smoke
10
null
null
1
null
null
null
12B
QAT
smoke
25
null
null
1
null
null
null
12B
QAT
smoke
15
null
null
0.9333
null
null
null
12B
QAT
smoke
35
null
null
0.6857
null
null
null
12B
QAT
core500
25
null
null
0.6
null
null
null
12B
QAT
core500
50
null
null
0.64
null
null
null
12B
QAT
core500
50
null
null
0.92
null
null
null
12B
QAT
core500
125
null
null
0.984
null
null
null
12B
QAT
core500
75
null
null
0.9467
null
null
null
12B
QAT
core500
175
null
null
0.4971
null
null
null
26B
PTQ
smoke
5
null
null
0.6
null
null
null
26B
PTQ
smoke
10
null
null
0.9
null
null
null
26B
PTQ
smoke
10
null
null
1
null
null
null
26B
PTQ
smoke
25
null
null
1
null
null
null
26B
PTQ
smoke
15
null
null
0.9333
null
null
null
26B
PTQ
smoke
35
null
null
0.6286
null
null
null
26B
PTQ
core500
25
null
null
0.48
null
null
null
26B
PTQ
core500
50
null
null
0.6
null
null
null
26B
PTQ
core500
50
null
null
0.92
null
null
null
26B
PTQ
core500
125
null
null
0.976
null
null
null
26B
PTQ
core500
75
null
null
0.96
null
null
null
26B
PTQ
core500
175
null
null
0.4971
null
null
null
26B
QAT
smoke
5
null
null
0.8
null
null
null
26B
QAT
smoke
10
null
null
0.8
null
null
null
26B
QAT
smoke
10
null
null
1
null
null
null
26B
QAT
smoke
25
null
null
1
null
null
null
26B
QAT
smoke
15
null
null
0.9333
null
null
null
26B
QAT
smoke
35
null
null
0.7143
null
null
null
26B
QAT
core500
25
null
null
0.6
null
null
null
26B
QAT
core500
50
null
null
0.58
null
null
null
26B
QAT
core500
50
null
null
0.92
null
null
null
26B
QAT
core500
125
null
null
0.984
null
null
null
26B
QAT
core500
75
null
null
0.96
null
null
null
26B
QAT
core500
175
null
null
0.5314
null
null
null
12B
null
core500
null
null
null
null
null
null
null
26B
null
core500
null
null
null
null
null
null
null
12B
null
smoke
null
null
null
null
null
null
null
26B
null
smoke
null
null
null
null
null
null
null
12B
null
core500
500
null
null
0.296
null
null
null
26B
null
core500
500
null
null
0.232
null
null
null
12B
null
smoke
100
null
null
0.32
null
null
null
26B
null
smoke
100
null
null
0.27
null
null
null

Gemma 4 Quantization-Regime Study (QAT Q4_0 vs PTQ Q4_K_M)

A same-day 2×2 study (quantization regime × released model pair) of governed routing quality for Gemma 4 12B IT (dense; a clean same-base pair) and a 26B-class MoE released pair on one RTX 5070 Ti, run under a fixed, bit-identical governance stack via Ollama. Companion/sequel to the Gemma 4 MTP Quality–Throughput Study (2026-08-09, DOI 10.5281/zenodo.21860461).

Author: Taiko Toeda, Independent Researcher ORCID: https://orcid.org/0009-0001-7267-0201 Version: 1.0.1 DOI: 10.5281/zenodo.21875513 (published 2026-08-11)

Result boundary

Aggregate route correctness spanned 0.738–0.760 on Core-500 (N=500 rows = 100 paraphrase stems × 5); no equivalence test was run, and one cross-model contrast (12B-PTQ vs 26B-PTQ) is nominally significant (p=0.035). The within-pair regime effects were individually marginal and oppositely signed — QAT −0.012 on the dense 12B (McNemar p=0.070) and +0.018 on the 26B released pair (p=0.078) — and the regime × pair interaction, an exploratory single-run endpoint, was +0.030 with a stem-clustered bootstrap 95% CI of [+0.010, +0.052] (task-iid CI [+0.008, +0.052]), concentrated in the volatile-current task family (crossed pattern). Cross-regime byte-identity of temperature-0 outputs was only 23–32 %, against same-configuration Smoke-100 repeats of 100/100 in all four cells (including the CPU-offloaded 26B-PTQ cell, across run windows) and a 100/100 num_ctx byte-identity control. Safety-critical failure rates were 0.000 in seven of eight runs, with a single over-verification event (0.008) in 26B-PTQ Core-500.

These results apply only to the measured GPU, runtime (Ollama 0.30.6), the released model artifacts named in ENVIRONMENT.md, one governance stack, and routing/abstention-style tasks. The 26B released pair confounds regime with a possible base revision (25.8B vs 25.2B), an expert configuration unverifiable from runtime metadata, and a CPU-offload compute path; only the 12B pair is a clean same-base, all-GPU contrast. All inferential results are exploratory (no preregistration, single run per cell, no multiplicity correction). See CLAIMS_MATRIX.md before reusing any claim.

Public artifact boundary

This public package excludes verbatim task prompts, generated output text, raw row-level request records, local paths, and the private internal archive. It retains the paper, processed measurements, per-family and paired statistics, environment and model digests, the deterministic analysis script, and SHA-256 hashes that identify every excluded raw artifact (data/RAW_ARTIFACT_HASHES.sha256). The evaluation harness is public at github.com/mobius-style/mmv (operate-fr-bench/); the OPERATE-FR suite items themselves are project-local and identified by hash in ENVIRONMENT.md.

Layout

PAPER.md — full study. EXECUTIVE_SUMMARY.md — condensed findings. CLAIMS_MATRIX.md — claim-by-claim evidence boundaries. REPRODUCE.md — commands and artifact map. ENVIRONMENT.md — pinned environment, model digests, suite hashes. data/ — processed tables and statistics. analysis/ — deterministic analysis script (Apache-2.0). Licensing: see LICENSE.md (split license).

Downloads last month
14