Dataset Viewer
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
data_key: string
subject_model: string
layer: int64
hookpoint: string
dataset: string
n_seqs: int64
seq_len: int64
d_in: int64
substrate: string
corpus_repo: string
corpus_seed: int64
builder: string
questions: list<item: string>
child 0, item: string
base_model: string
run_ref: string
wall_seconds: double
misaligned_frac: double
hs_all: list<item: int64>
child 0, item: int64
d_model: int64
n_rollouts: int64
adapter: string
to
{'base_model': Value('string'), 'adapter': Value('string'), 'run_ref': Value('string'), 'n_rollouts': Value('int64'), 'd_model': Value('int64'), 'hs_all': List(Value('int64')), 'misaligned_frac': Value('float64'), 'questions': List(Value('string')), 'wall_seconds': Value('float64')}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
return get_rows(
dataset=dataset,
...<4 lines>...
column_names=column_names,
)
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 129, in get_rows
rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
File "/src/services/worker/src/worker/utils.py", line 489, in safe_iter
yield from ds.decode(False) if ds.features else ds
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2818, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2355, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2380, in _iter_arrow
for key, pa_table in self.ex_iterable._iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_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
data_key: string
subject_model: string
layer: int64
hookpoint: string
dataset: string
n_seqs: int64
seq_len: int64
d_in: int64
substrate: string
corpus_repo: string
corpus_seed: int64
builder: string
questions: list<item: string>
child 0, item: string
base_model: string
run_ref: string
wall_seconds: double
misaligned_frac: double
hs_all: list<item: int64>
child 0, item: int64
d_model: int64
n_rollouts: int64
adapter: string
to
{'base_model': Value('string'), 'adapter': Value('string'), 'run_ref': Value('string'), 'n_rollouts': Value('int64'), 'd_model': Value('int64'), 'hs_all': List(Value('int64')), 'misaligned_frac': Value('float64'), 'questions': List(Value('string')), 'wall_seconds': Value('float64')}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
temp-xc btk-sprint artifacts
Artifacts from the 2026-07-26/27 "batch_topk" sprint (branch
dmitry-btk-txc-sprint of chainik1125/temp_xc): the paper-arch
activation 2×2 (per-window vs batch-pooled selection × ReLU placement)
and the §5.3 EM window-scaling sweep.
Contents
em_checkpoints/<train_key>/— trained EM dictionaries (model.safetensors+config.json), Qwen2.5-7B-Instruct L15 medical-organism task, d_sae=32768, canonical 25k-step training. Arms:txc_base(paper composite) andtxc_base_btkonlyat T ∈ {1, 4, 6, 16}, seed 42. Keys map viamanifests/*.jsonl.em_data_cache/<data_key>/— BASE-forward training activation cache forqwen_2_5_7b_instruct_medical_l15(6000×128, fp16).conv_depth_caches/em_medical/— cohort eval caches (hs16.npy, labels, lens).judge_outputs.jsonlis deliberately excluded (raw misaligned-organism generations; kept on private volumes).- Leaderboard rows and analysis live in git on the branch
(
results/leaderboard.jsonl,summary.md,plots/btk_rerun/).
Headline result: EM detection (pr_auc_S16) under the paper arch is an
inverted-U in window size T (0.478 → 0.581 → 0.598 → 0.481 over
T ∈ {1,4,6,16}); the btk-only arm holds the T=16 tail (0.578) at
5–49× realized eval density. Full context: summary.md on the branch.
- Downloads last month
- 45