Success Conditioning as Policy Improvement: The Optimization Problem Solved by Imitating Success
Paper • 2601.18175 • Published
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
python: string
executable: string
platform: string
processor: string
packages: struct<numpy: string, scipy: string, pandas: string, matplotlib: string, trackio: string, huggingfac (... 14 chars omitted)
child 0, numpy: string
child 1, scipy: string
child 2, pandas: string
child 3, matplotlib: string
child 4, trackio: string
child 5, huggingface_hub: string
kind: string
seed: int64
horizon: int64
n_states: int64
instance_id: string
terminal_signal: list<item: double>
child 0, item: double
policy0: list<item: list<item: list<item: double>>>
child 0, item: list<item: list<item: double>>
child 0, item: list<item: double>
child 0, item: double
n_actions: int64
transition: list<item: list<item: list<item: list<item: double>>>>
child 0, item: list<item: list<item: list<item: double>>>
child 0, item: list<item: list<item: double>>
child 0, item: list<item: double>
child 0, item: double
mu: list<item: double>
child 0, item: double
to
{'instance_id': Value('string'), 'seed': Value('int64'), 'kind': Value('string'), 'horizon': Value('int64'), 'n_states': Value('int64'), 'n_actions': Value('int64'), 'mu': List(Value('float64')), 'terminal_signal': List(Value('float64')), 'policy0': List(List(List(Value('float64')))), 'transition': List(List(List(List(Value('float64')))))}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 147, 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 127, 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 478, 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
python: string
executable: string
platform: string
processor: string
packages: struct<numpy: string, scipy: string, pandas: string, matplotlib: string, trackio: string, huggingfac (... 14 chars omitted)
child 0, numpy: string
child 1, scipy: string
child 2, pandas: string
child 3, matplotlib: string
child 4, trackio: string
child 5, huggingface_hub: string
kind: string
seed: int64
horizon: int64
n_states: int64
instance_id: string
terminal_signal: list<item: double>
child 0, item: double
policy0: list<item: list<item: list<item: double>>>
child 0, item: list<item: list<item: double>>
child 0, item: list<item: double>
child 0, item: double
n_actions: int64
transition: list<item: list<item: list<item: list<item: double>>>>
child 0, item: list<item: list<item: list<item: double>>>
child 0, item: list<item: list<item: double>>
child 0, item: list<item: double>
child 0, item: double
mu: list<item: double>
child 0, item: double
to
{'instance_id': Value('string'), 'seed': Value('int64'), 'kind': Value('string'), 'horizon': Value('int64'), 'n_states': Value('int64'), 'n_actions': Value('int64'), 'mu': List(Value('float64')), 'terminal_signal': List(Value('float64')), 'policy0': List(List(List(Value('float64')))), 'transition': List(List(List(List(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.
Success-Conditioning as Policy Improvement — ICML 2026 reproduction
This artifact independently verifies the exact finite-tabular claims in Daniel Russo's Success-Conditioning as Policy Improvement: The Optimization Problem Solved by Imitating Success (arXiv:2601.18175, OpenReview:FEmXFeqYNZ).
The reproduction is deterministic, CPU-bound, and intentionally does not claim to be a deep-RL benchmark. It uses exact dynamic programming, an independently parameterized SciPy SLSQP solve, exhaustive trajectory conditioning on small instances, and adversarial edge cases. The substantive cloud run is submitted on a Hugging Face GPU Job for challenge compliance; the GPU is not scientifically needed by the tabular workload.
Reproduce locally
uv sync --extra dev
uv run pytest
uv run success-conditioning-repro --config configs/full.json --output outputs/full
All seeds, complete MDP tensors, per-state results, solver restarts, figures, environment details, and checksums are written beneath the selected output directory.
- Downloads last month
- 17