Strategic Candidacy in Generative AI Arenas
Paper • 2603.26891 • Published
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
mode: string
import_core_ok: bool
n_checks: int64
n_passed: int64
all_passed: bool
total_wall_seconds: double
checks: struct<win_prob_matrix: struct<name: string, passed: bool, detail: struct<shape: list<item: int64>, (... 1057 chars omitted)
child 0, win_prob_matrix: struct<name: string, passed: bool, detail: struct<shape: list<item: int64>, dtype: string>, error: n (... 26 chars omitted)
child 0, name: string
child 1, passed: bool
child 2, detail: struct<shape: list<item: int64>, dtype: string>
child 0, shape: list<item: int64>
child 0, item: int64
child 1, dtype: string
child 3, error: null
child 4, wall_seconds: double
child 1, draw_winrate_matrix: struct<name: string, passed: bool, detail: struct<shape: list<item: int64>>, error: null, wall_secon (... 11 chars omitted)
child 0, name: string
child 1, passed: bool
child 2, detail: struct<shape: list<item: int64>>
child 0, shape: list<item: int64>
child 0, item: int64
child 3, error: null
child 4, wall_seconds: double
child 2, draw_winrate_matrix_with_clones: struct<name: string, passed: bool, detail: struct<shape: list<item: int64>, n_clones: int64>, error: (... 28 chars omitted)
child 0, name: string
child 1, passed: bool
child 2, detail: struct<shape: list<item: int64>, n_clones: int64>
child 0, shape: list<item: int64>
child 0, item: int64
child
...
ild 1, passed: bool
child 2, detail: struct<shape: list<item: int64>, max_abs_err_vs_true: double>
child 0, shape: list<item: int64>
child 0, item: int64
child 1, max_abs_err_vs_true: double
child 3, error: null
child 4, wall_seconds: double
child 5, rank_of: struct<name: string, passed: bool, detail: struct<ok: bool>, error: null, wall_seconds: double>
child 0, name: string
child 1, passed: bool
child 2, detail: struct<ok: bool>
child 0, ok: bool
child 3, error: null
child 4, wall_seconds: double
child 6, yrwr_correct_running_min_equiv: struct<name: string, passed: bool, detail: struct<max_abs_diff_vs_running_min: double>, error: null, (... 22 chars omitted)
child 0, name: string
child 1, passed: bool
child 2, detail: struct<max_abs_diff_vs_running_min: double>
child 0, max_abs_diff_vs_running_min: double
child 3, error: null
child 4, wall_seconds: double
child 7, seeded_rng_determinism: struct<name: string, passed: bool, detail: struct<deterministic: bool, distinct_for_distinct_keys: b (... 40 chars omitted)
child 0, name: string
child 1, passed: bool
child 2, detail: struct<deterministic: bool, distinct_for_distinct_keys: bool>
child 0, deterministic: bool
child 1, distinct_for_distinct_keys: bool
child 3, error: null
child 4, wall_seconds: double
exp02: string
exp01: string
exp00: string
to
{'exp00': Value('string'), 'exp01': Value('string'), 'exp02': Value('string')}
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
mode: string
import_core_ok: bool
n_checks: int64
n_passed: int64
all_passed: bool
total_wall_seconds: double
checks: struct<win_prob_matrix: struct<name: string, passed: bool, detail: struct<shape: list<item: int64>, (... 1057 chars omitted)
child 0, win_prob_matrix: struct<name: string, passed: bool, detail: struct<shape: list<item: int64>, dtype: string>, error: n (... 26 chars omitted)
child 0, name: string
child 1, passed: bool
child 2, detail: struct<shape: list<item: int64>, dtype: string>
child 0, shape: list<item: int64>
child 0, item: int64
child 1, dtype: string
child 3, error: null
child 4, wall_seconds: double
child 1, draw_winrate_matrix: struct<name: string, passed: bool, detail: struct<shape: list<item: int64>>, error: null, wall_secon (... 11 chars omitted)
child 0, name: string
child 1, passed: bool
child 2, detail: struct<shape: list<item: int64>>
child 0, shape: list<item: int64>
child 0, item: int64
child 3, error: null
child 4, wall_seconds: double
child 2, draw_winrate_matrix_with_clones: struct<name: string, passed: bool, detail: struct<shape: list<item: int64>, n_clones: int64>, error: (... 28 chars omitted)
child 0, name: string
child 1, passed: bool
child 2, detail: struct<shape: list<item: int64>, n_clones: int64>
child 0, shape: list<item: int64>
child 0, item: int64
child
...
ild 1, passed: bool
child 2, detail: struct<shape: list<item: int64>, max_abs_err_vs_true: double>
child 0, shape: list<item: int64>
child 0, item: int64
child 1, max_abs_err_vs_true: double
child 3, error: null
child 4, wall_seconds: double
child 5, rank_of: struct<name: string, passed: bool, detail: struct<ok: bool>, error: null, wall_seconds: double>
child 0, name: string
child 1, passed: bool
child 2, detail: struct<ok: bool>
child 0, ok: bool
child 3, error: null
child 4, wall_seconds: double
child 6, yrwr_correct_running_min_equiv: struct<name: string, passed: bool, detail: struct<max_abs_diff_vs_running_min: double>, error: null, (... 22 chars omitted)
child 0, name: string
child 1, passed: bool
child 2, detail: struct<max_abs_diff_vs_running_min: double>
child 0, max_abs_diff_vs_running_min: double
child 3, error: null
child 4, wall_seconds: double
child 7, seeded_rng_determinism: struct<name: string, passed: bool, detail: struct<deterministic: bool, distinct_for_distinct_keys: b (... 40 chars omitted)
child 0, name: string
child 1, passed: bool
child 2, detail: struct<deterministic: bool, distinct_for_distinct_keys: bool>
child 0, deterministic: bool
child 1, distinct_for_distinct_keys: bool
child 3, error: null
child 4, wall_seconds: double
exp02: string
exp01: string
exp00: string
to
{'exp00': Value('string'), 'exp01': Value('string'), 'exp02': Value('string')}
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.
Full reproduction artifacts for the ICML-2026-agent-repro reproduction of "Strategic Candidacy in Generative AI Arenas" (arXiv:2603.26891, OpenReview K1mAvyHhKU). Published logbook: https://huggingface.co/spaces/JG1310/repro-strategic-candidacy-in-generative-ai-arenas
DERIVATIONS.md — re-derivations of the 5 theorems/lemmas (D1–D5) the two claims rest on, each paired with an executable numerical check.BRIEF_WRITER.md — claim → evidence map used to write the logbook.STATE.md — experiment plan / ledger.specs/ — experiment specs (exp00 core library, exp01 clone simulation, exp02 theory checks), each with an "Acceptance gate" section.scripts/ — implementation (core.py shared library; exp00_core.py, exp01_clone_simulation.py, exp02_theory_checks.py).results/*.json — full-scale (and toy) outputs; GATE_REPORT.txt — structural fidelity gate output (gates.py, schema/shape/provenance checks only, no outcome thresholds).logs/*.log — run logs (stderr progress, wall-clock).paper.txt — plain-text extraction of the paper, used for verbatim quotes.python3 scripts/exp00_core.py --full
python3 scripts/exp01_clone_simulation.py --full # real Arena data (lmarena-ai/leaderboard-dataset); ~7h/8 cores
python3 scripts/exp02_theory_checks.py --full # data-free mechanism checks; ~4h/4 cores
python3 gates.py exp00 --full --report
python3 gates.py exp01 --full --report
python3 gates.py exp02 --full --report