Dataset Preview
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 12 new columns ({'eval_seed', 'checkpoint_final_delta', 'seed', 'eval_batches', 'in_loop_baseline_final', 'checkpoint_baseline_final', 'condition_path', 'baseline_path', 'in_loop_final_delta', 'checkpoint_minus_in_loop_delta', 'checkpoint_condition_final', 'in_loop_condition_final'}) and 8 missing columns ({'delta_shift_mean', 'delta_shift_std', 'checkpoint_final_wins', 'checkpoint_final_delta_mean', 'in_loop_final_delta_mean', 'in_loop_final_wins', 'checkpoint_final_delta_std', 'n'}).
This happened while the csv dataset builder was generating data using
hf://datasets/cuber12/budgeted-guardrails-mlp-channel-interventions/reports/post_training_eval/claim_hardening_complete_pairs/checkpoint_eval_paired_deltas.csv (at revision e2140caed0a9357bea5e2026c18271f117057fdf), ['hf://datasets/cuber12/budgeted-guardrails-mlp-channel-interventions@e2140caed0a9357bea5e2026c18271f117057fdf/reports/post_training_eval/claim_hardening_complete_pairs/checkpoint_eval_condition_summaries.csv', 'hf://datasets/cuber12/budgeted-guardrails-mlp-channel-interventions@e2140caed0a9357bea5e2026c18271f117057fdf/reports/post_training_eval/claim_hardening_complete_pairs/checkpoint_eval_paired_deltas.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
group: string
condition: string
seed: int64
eval_seed: int64
eval_batches: int64
checkpoint_condition_final: double
checkpoint_baseline_final: double
checkpoint_final_delta: double
in_loop_condition_final: double
in_loop_baseline_final: double
in_loop_final_delta: double
checkpoint_minus_in_loop_delta: double
condition_path: string
baseline_path: string
-- schema metadata --
pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 2142
to
{'group': Value('string'), 'condition': Value('string'), 'n': Value('int64'), 'checkpoint_final_delta_mean': Value('float64'), 'checkpoint_final_delta_std': Value('float64'), 'checkpoint_final_wins': Value('int64'), 'in_loop_final_delta_mean': Value('float64'), 'in_loop_final_wins': Value('int64'), 'delta_shift_mean': Value('float64'), 'delta_shift_std': 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 12 new columns ({'eval_seed', 'checkpoint_final_delta', 'seed', 'eval_batches', 'in_loop_baseline_final', 'checkpoint_baseline_final', 'condition_path', 'baseline_path', 'in_loop_final_delta', 'checkpoint_minus_in_loop_delta', 'checkpoint_condition_final', 'in_loop_condition_final'}) and 8 missing columns ({'delta_shift_mean', 'delta_shift_std', 'checkpoint_final_wins', 'checkpoint_final_delta_mean', 'in_loop_final_delta_mean', 'in_loop_final_wins', 'checkpoint_final_delta_std', 'n'}).
This happened while the csv dataset builder was generating data using
hf://datasets/cuber12/budgeted-guardrails-mlp-channel-interventions/reports/post_training_eval/claim_hardening_complete_pairs/checkpoint_eval_paired_deltas.csv (at revision e2140caed0a9357bea5e2026c18271f117057fdf), ['hf://datasets/cuber12/budgeted-guardrails-mlp-channel-interventions@e2140caed0a9357bea5e2026c18271f117057fdf/reports/post_training_eval/claim_hardening_complete_pairs/checkpoint_eval_condition_summaries.csv', 'hf://datasets/cuber12/budgeted-guardrails-mlp-channel-interventions@e2140caed0a9357bea5e2026c18271f117057fdf/reports/post_training_eval/claim_hardening_complete_pairs/checkpoint_eval_paired_deltas.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.
group string | condition string | n int64 | checkpoint_final_delta_mean float64 | checkpoint_final_delta_std float64 | checkpoint_final_wins int64 | in_loop_final_delta_mean float64 | in_loop_final_wins int64 | delta_shift_mean float64 | delta_shift_std float64 |
|---|---|---|---|---|---|---|---|---|---|
tinyshakespeare_relu_l4_8k | blind_late_gate_frac_0p01 | 3 | 0.000742 | 0.002974 | 1 | -0.004178 | 1 | 0.00492 | 0.015951 |
tinyshakespeare_relu_l4_8k | dropout_0p02 | 3 | 0.007547 | 0.00518 | 0 | 0.008453 | 2 | -0.000906 | 0.022093 |
tinyshakespeare_relu_l4_8k | early_budget_gate_frac_0p01 | 3 | -0.00008 | 0.003622 | 1 | -0.006558 | 2 | 0.006479 | 0.015756 |
tinyshakespeare_relu_l4_8k | randommatched_gate_only | 3 | -0.003214 | 0.008651 | 2 | -0.009185 | 2 | 0.005971 | 0.032559 |
tinyshakespeare_relu_l4_8k | regret_gate_frac_0p01 | 3 | -0.000809 | 0.007502 | 1 | 0.003628 | 1 | -0.004437 | 0.005792 |
wikitext2_relu_l4_16k | blind_late_gate_frac_0p01 | 3 | 0.000986 | 0.003181 | 2 | -0.001508 | 1 | 0.002494 | 0.011953 |
wikitext2_relu_l4_16k | dropout_0p02 | 3 | 0.008355 | 0.001699 | 0 | -0.001523 | 2 | 0.009878 | 0.02003 |
wikitext2_relu_l4_16k | early_budget_gate_frac_0p01 | 3 | 0.000304 | 0.002449 | 2 | -0.004048 | 1 | 0.004352 | 0.013393 |
wikitext2_relu_l4_16k | randommatched_gate_only | 3 | -0.000563 | 0.001564 | 2 | -0.007302 | 2 | 0.006739 | 0.028884 |
wikitext2_relu_l4_16k | regret_gate_frac_0p01 | 3 | 0.001316 | 0.000221 | 0 | 0.000139 | 1 | 0.001178 | 0.015832 |
wikitext2_relu_l8d256_4k | blind_late_gate_frac_0p01 | 3 | 0.001892 | 0.011566 | 1 | 0.036174 | 0 | -0.034281 | 0.017012 |
wikitext2_relu_l8d256_4k | dropout_0p02 | 3 | 0.01939 | 0.013253 | 0 | 0.028162 | 0 | -0.008772 | 0.016727 |
wikitext2_relu_l8d256_4k | early_budget_gate_frac_0p01 | 3 | 0.004043 | 0.012771 | 1 | 0.034586 | 1 | -0.030543 | 0.021479 |
wikitext2_relu_l8d256_4k | randommatched_gate_only | 3 | 0.003114 | 0.013533 | 1 | 0.023836 | 1 | -0.020722 | 0.033659 |
wikitext2_relu_l8d256_4k | regret_gate_frac_0p01 | 3 | 0.006243 | 0.004644 | 0 | 0.007402 | 1 | -0.001159 | 0.011391 |
wikitext2_squared_relu_l4_8k | blind_late_gate_frac_0p01 | 3 | 0.000054 | 0.004523 | 1 | -0.001601 | 1 | 0.001655 | 0.016697 |
wikitext2_squared_relu_l4_8k | dropout_0p02 | 3 | 0.011938 | 0.000797 | 0 | -0.006541 | 2 | 0.018479 | 0.044139 |
wikitext2_squared_relu_l4_8k | early_budget_gate_frac_0p01 | 3 | -0.00091 | 0.002913 | 2 | -0.005239 | 1 | 0.004329 | 0.017376 |
wikitext2_squared_relu_l4_8k | randommatched_gate_only | 3 | 0.001085 | 0.004598 | 1 | -0.01203 | 2 | 0.013115 | 0.047061 |
wikitext2_squared_relu_l4_8k | regret_gate_frac_0p01 | 3 | -0.000199 | 0.004993 | 2 | -0.006028 | 2 | 0.005829 | 0.01317 |
tinyshakespeare_relu_l4_8k | blind_late_gate_frac_0p01 | null | null | null | null | null | null | null | null |
tinyshakespeare_relu_l4_8k | blind_late_gate_frac_0p01 | null | null | null | null | null | null | null | null |
tinyshakespeare_relu_l4_8k | blind_late_gate_frac_0p01 | null | null | null | null | null | null | null | null |
tinyshakespeare_relu_l4_8k | dropout_0p02 | null | null | null | null | null | null | null | null |
tinyshakespeare_relu_l4_8k | dropout_0p02 | null | null | null | null | null | null | null | null |
tinyshakespeare_relu_l4_8k | dropout_0p02 | null | null | null | null | null | null | null | null |
tinyshakespeare_relu_l4_8k | early_budget_gate_frac_0p01 | null | null | null | null | null | null | null | null |
tinyshakespeare_relu_l4_8k | early_budget_gate_frac_0p01 | null | null | null | null | null | null | null | null |
tinyshakespeare_relu_l4_8k | early_budget_gate_frac_0p01 | null | null | null | null | null | null | null | null |
tinyshakespeare_relu_l4_8k | randommatched_gate_only | null | null | null | null | null | null | null | null |
tinyshakespeare_relu_l4_8k | randommatched_gate_only | null | null | null | null | null | null | null | null |
tinyshakespeare_relu_l4_8k | randommatched_gate_only | null | null | null | null | null | null | null | null |
tinyshakespeare_relu_l4_8k | regret_gate_frac_0p01 | null | null | null | null | null | null | null | null |
tinyshakespeare_relu_l4_8k | regret_gate_frac_0p01 | null | null | null | null | null | null | null | null |
tinyshakespeare_relu_l4_8k | regret_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l4_16k | blind_late_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l4_16k | blind_late_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l4_16k | blind_late_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l4_16k | dropout_0p02 | null | null | null | null | null | null | null | null |
wikitext2_relu_l4_16k | dropout_0p02 | null | null | null | null | null | null | null | null |
wikitext2_relu_l4_16k | dropout_0p02 | null | null | null | null | null | null | null | null |
wikitext2_relu_l4_16k | early_budget_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l4_16k | early_budget_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l4_16k | early_budget_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l4_16k | randommatched_gate_only | null | null | null | null | null | null | null | null |
wikitext2_relu_l4_16k | randommatched_gate_only | null | null | null | null | null | null | null | null |
wikitext2_relu_l4_16k | randommatched_gate_only | null | null | null | null | null | null | null | null |
wikitext2_relu_l4_16k | regret_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l4_16k | regret_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l4_16k | regret_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l8d256_4k | blind_late_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l8d256_4k | blind_late_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l8d256_4k | blind_late_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l8d256_4k | dropout_0p02 | null | null | null | null | null | null | null | null |
wikitext2_relu_l8d256_4k | dropout_0p02 | null | null | null | null | null | null | null | null |
wikitext2_relu_l8d256_4k | dropout_0p02 | null | null | null | null | null | null | null | null |
wikitext2_relu_l8d256_4k | early_budget_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l8d256_4k | early_budget_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l8d256_4k | early_budget_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l8d256_4k | randommatched_gate_only | null | null | null | null | null | null | null | null |
wikitext2_relu_l8d256_4k | randommatched_gate_only | null | null | null | null | null | null | null | null |
wikitext2_relu_l8d256_4k | randommatched_gate_only | null | null | null | null | null | null | null | null |
wikitext2_relu_l8d256_4k | regret_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l8d256_4k | regret_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_relu_l8d256_4k | regret_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_squared_relu_l4_8k | blind_late_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_squared_relu_l4_8k | blind_late_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_squared_relu_l4_8k | blind_late_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_squared_relu_l4_8k | dropout_0p02 | null | null | null | null | null | null | null | null |
wikitext2_squared_relu_l4_8k | dropout_0p02 | null | null | null | null | null | null | null | null |
wikitext2_squared_relu_l4_8k | dropout_0p02 | null | null | null | null | null | null | null | null |
wikitext2_squared_relu_l4_8k | early_budget_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_squared_relu_l4_8k | early_budget_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_squared_relu_l4_8k | early_budget_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_squared_relu_l4_8k | randommatched_gate_only | null | null | null | null | null | null | null | null |
wikitext2_squared_relu_l4_8k | randommatched_gate_only | null | null | null | null | null | null | null | null |
wikitext2_squared_relu_l4_8k | randommatched_gate_only | null | null | null | null | null | null | null | null |
wikitext2_squared_relu_l4_8k | regret_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_squared_relu_l4_8k | regret_gate_frac_0p01 | null | null | null | null | null | null | null | null |
wikitext2_squared_relu_l4_8k | regret_gate_frac_0p01 | null | null | null | null | null | null | null | null |
MLP Channel Guardrails
This repository accompanies the Zenodo preprint
10.5281/zenodo.21003379,
Budgeted Guardrails for MLP Channel Interventions in Small Transformer
Language Models.
The study evaluates training-time MLP-channel interventions in small causal Transformer language models. The supported claim is narrow: in the tested small-model settings, low-amplitude MLP channel gates with strict event budgets reduce trajectory damage relative to fixed event gating and dropout while keeping the trained model dense at inference time.
The project no longer claims that targeted dormant-channel rejuvenation works. Matched-random controls are as strong or stronger than dormant-targeted interventions in the current evidence.
Release Artifacts
This Hugging Face dataset repository is a public mirror of the Zenodo preprint release package.
Primary release files:
output/pdf/budgeted_guardrails_mlp_channel_interventions.pdfoutput/zenodo/budgeted-guardrails-mlp-channel-interventions-20260617.zippaper/budgeted_guardrails_mlp_channel_interventions.texpaper/references.bibpaper/reproducibility_manifest.mdpaper/zenodo_metadata.json
Recommended citation metadata and artifact details are in
README_ZENODO.md.
Repository Contents
The release package includes:
- Small GPT-style causal Transformer training code.
- MLP-channel telemetry and channel-gating intervention code.
- Launch, aggregation, statistical testing, and figure-generation scripts.
- Generated diagnostics and manuscript tables under
reports/andpaper/tables/. - Public bounded text-data caches and metadata under
data/raw/. - Dependency-free tests under
tests/.
Paper reproducibility details are in
paper/reproducibility_manifest.md.
Training experiments must use MPS only. If MPS is unavailable, stop before
model creation. CPU, CUDA, and PYTORCH_ENABLE_MPS_FALLBACK=1 runs are invalid
for this project.
Verification
Dependency-free tests:
PYTHONPATH=src python -m unittest discover -s tests -v
MPS gate:
PYTHONPATH=src .venv/bin/python scripts/check_mps.py
Paper consistency check for result-table, claim-language, bibliography, figure-reference, CI-artifact, abstract-length, and section-order drift:
PYTHONPATH=src .venv/bin/python scripts/check_paper_consistency.py
First smoke run, only after the MPS gate passes:
PYTHONPATH=src .venv/bin/python -m dormant_rejuvenation.train \
--run-name smoke_telemetry \
--out-dir runs/exploratory \
--steps 50 \
--eval-every 25 \
--telemetry-every 5
Bounded corpus sample download:
python scripts/download_text_sample.py \
--name tinystories_sample \
--url https://huggingface.co/datasets/roneneldan/TinyStories/resolve/main/TinyStories-train.txt \
--out data/raw/tinystories_sample.txt \
--max-bytes 8000000
- Downloads last month
- 147