Evolution Strategies with LLM Novelty
Collection
Preregistered full-weight ES, correctness-gated novelty, controls and reports. • 3 items • Updated
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: TypeError
Message: Couldn't cast array of type
list<item: struct<problem_id: string, prompt: string, response: string, numbers: list<item: int64>, target: int64, generated_tokens: int64, token_ids: list<item: int64>, truncated: bool, batch_seconds: double, reward: double, correct: bool, valid: bool, expression: string, structure: string, sample: int64>>
to
{'test': List({'problem_id': Value('string'), 'prompt': Value('string'), 'response': Value('string'), 'numbers': List(Value('int64')), 'target': Value('int64'), 'generated_tokens': Value('int64'), 'token_ids': List(Value('int64')), 'truncated': Value('bool'), 'batch_seconds': Value('float64'), 'reward': Value('float64'), 'correct': Value('bool'), 'valid': Value('bool'), 'expression': Value('string'), 'structure': Value('string'), 'sample': Value('null')}), 'hard': List({'problem_id': Value('string'), 'prompt': Value('string'), 'response': Value('string'), 'numbers': List(Value('int64')), 'target': Value('int64'), 'generated_tokens': Value('int64'), 'token_ids': List(Value('int64')), 'truncated': Value('bool'), 'batch_seconds': Value('float64'), 'reward': Value('float64'), 'correct': Value('bool'), 'valid': Value('bool'), 'expression': Value('string'), 'structure': Value('string'), 'sample': Value('null')}), 'probes': List({'problem_id': Value('string'), 'prompt': Value('string'), 'response': Value('string'), 'numbers': List(Value('int64')), 'target': Value('int64'), 'generated_tokens': Value('int64'), 'token_ids': List(Value('int64')), 'truncated': Value('bool'), 'batch_seconds': Value('float64'), 'reward': Value('float64'), 'correct': Value('bool'), 'valid': Value('bool'), 'expression': Value('string'), 'structure': Value('string'), 'sample': Value('null')}), 'sampled': List({'problem_id': Value('string'), 'prompt': Value('string'), 'response': Value('string'), 'numbers': List(Value('int64')), 'target': Value('int64'), 'generated_tokens': Value('int64'), 'token_ids': List(Value('int64')), 'truncated': Value('bool'), 'batch_seconds': Value('float64'), 'reward': Value('float64'), 'correct': Value('bool'), 'valid': Value('bool'), 'expression': Value('string'), 'structure': Value('string'), 'sample': Value('int64')})}
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 291, in _generate_tables
pa_table = paj.read_json(
io.BytesIO(batch), read_options=paj.ReadOptions(block_size=block_size)
)
File "pyarrow/_json.pyx", line 364, in pyarrow._json.read_json
File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
raise convert_status(status)
pyarrow.lib.ArrowInvalid: JSON parse error: Missing a name for object member. in row 0
During handling of the above exception, another exception occurred:
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 544, 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 3052, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, 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 558, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, 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 357, in _generate_tables
yield Key(shard_idx, 0), self._cast_table(pa_table)
~~~~~~~~~~~~~~~~^^^^^^^^^^
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 2381, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2315, in cast_table_to_schema
cast_array_to_feature(
~~~~~~~~~~~~~~~~~~~~~^
table[name] if name in table_column_names else pa.array([None] * len(table), type=schema.field(name).type),
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
feature,
^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1861, in wrapper
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
~~~~^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2161, in cast_array_to_feature
raise TypeError(f"Couldn't cast array of type\n{_short_str(array.type)}\nto\n{_short_str(feature)}")
TypeError: Couldn't cast array of type
list<item: struct<problem_id: string, prompt: string, response: string, numbers: list<item: int64>, target: int64, generated_tokens: int64, token_ids: list<item: int64>, truncated: bool, batch_seconds: double, reward: double, correct: bool, valid: bool, expression: string, structure: string, sample: int64>>
to
{'test': List({'problem_id': Value('string'), 'prompt': Value('string'), 'response': Value('string'), 'numbers': List(Value('int64')), 'target': Value('int64'), 'generated_tokens': Value('int64'), 'token_ids': List(Value('int64')), 'truncated': Value('bool'), 'batch_seconds': Value('float64'), 'reward': Value('float64'), 'correct': Value('bool'), 'valid': Value('bool'), 'expression': Value('string'), 'structure': Value('string'), 'sample': Value('null')}), 'hard': List({'problem_id': Value('string'), 'prompt': Value('string'), 'response': Value('string'), 'numbers': List(Value('int64')), 'target': Value('int64'), 'generated_tokens': Value('int64'), 'token_ids': List(Value('int64')), 'truncated': Value('bool'), 'batch_seconds': Value('float64'), 'reward': Value('float64'), 'correct': Value('bool'), 'valid': Value('bool'), 'expression': Value('string'), 'structure': Value('string'), 'sample': Value('null')}), 'probes': List({'problem_id': Value('string'), 'prompt': Value('string'), 'response': Value('string'), 'numbers': List(Value('int64')), 'target': Value('int64'), 'generated_tokens': Value('int64'), 'token_ids': List(Value('int64')), 'truncated': Value('bool'), 'batch_seconds': Value('float64'), 'reward': Value('float64'), 'correct': Value('bool'), 'valid': Value('bool'), 'expression': Value('string'), 'structure': Value('string'), 'sample': Value('null')}), 'sampled': List({'problem_id': Value('string'), 'prompt': Value('string'), 'response': Value('string'), 'numbers': List(Value('int64')), 'target': Value('int64'), 'generated_tokens': Value('int64'), 'token_ids': List(Value('int64')), 'truncated': Value('bool'), 'batch_seconds': Value('float64'), 'reward': Value('float64'), 'correct': Value('bool'), 'valid': Value('bool'), 'expression': Value('string'), 'structure': Value('string'), 'sample': Value('int64')})}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.
Novelty ES pilot evidence
The directional improvement hypothesis was not supported in this seed: main accuracy did not exceed reward-only ES.
| Condition | Test accuracy | 95% interval | Hard accuracy | pass@1 | pass@8 | Status |
|---|---|---|---|---|---|---|
| original | 0.29% | 0.10% - 0.54% | 0.10% | 0.12% | 0.89% | complete |
| main | 1.12% | 0.68% - 1.61% | 0.00% | 0.27% | 2.06% | complete |
| reward_only | 1.12% | 0.68% - 1.61% | 0.10% | 0.22% | 1.67% | complete |
Problem uncertainty conditional on one training seed; excludes seed variability.
This repository contains raw evaluations, timestamped generation traces, configuration and code provenance, initialization and recovery audits, and the final resumable FP32 training checkpoints. Inference-ready BF16 models are separate items in the same Hugging Face collection. The checkpoints contain master tensors, RNG states, archive, cursor and the original absolute deadline. They preserve the completed run; loading them does not grant a new training allowance. Use the project's weights-only checkpoint loader in the recorded environment. The SHA-256 manifest covers the explicitly uploaded evidence files. The final upload receipt records immutable Hub commit revisions.
The source, synthetic datasets and their hashes, frozen configuration, dependency lock, detailed report and PDF are at https://github.com/BurnyCoder/evolutionary-strategies-novelty-search-llm-training . Run uv run novelty-es report --run-dir PATH from that checkout to regenerate the report from these artifacts. The method adapts ExTra and ES at Scale; it is not a reproduction. Amendment A preceded all training updates but followed calibration and partial original-model evaluation. This single-seed arithmetic study provides no evidence about general language quality or seed robustness.
- Downloads last month
- 51