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
instance_id: string
scenario: string
level: int64
seed: int64
payload: struct<space: struct<issues: list<item: struct<name: string, options: list<item: string>>>>, sheets: (... 823 chars omitted)
child 0, space: struct<issues: list<item: struct<name: string, options: list<item: string>>>>
child 0, issues: list<item: struct<name: string, options: list<item: string>>>
child 0, item: struct<name: string, options: list<item: string>>
child 0, name: string
child 1, options: list<item: string>
child 0, item: string
child 1, sheets: list<item: struct<agent: string, values: list<item: list<item: double>>, threshold: double>>
child 0, item: struct<agent: string, values: list<item: list<item: double>>, threshold: double>
child 0, agent: string
child 1, values: list<item: list<item: double>>
child 0, item: list<item: double>
child 0, item: double
child 2, threshold: double
child 2, rounds: int64
child 3, info: string
child 4, chat: bool
child 5, proposer: int64
child 6, veto: int64
child 7, min_accept: int64
child 8, discount: double
child 9, breakdown_risk: double
child 10, constraint: null
child 11, meta: struct<generator: string, seed: int64, issue_types: list<item: string>, option_perms: list<item: lis (... 471 chars omitted)
child 0, generator: string
child 1, seed: int64
child 2, issue_types: list<item: string>
...
ist<item: struct<instance_id: string, generator_seed: int64, level: int64, sha256: string, content_ (... 334 chars omitted)
child 0, item: struct<instance_id: string, generator_seed: int64, level: int64, sha256: string, content_sha256: str (... 322 chars omitted)
child 0, instance_id: string
child 1, generator_seed: int64
child 2, level: int64
child 3, sha256: string
child 4, content_sha256: string
child 5, difficulty_scalar: double
child 6, difficulty_tags: list<item: string>
child 0, item: string
child 7, difficulty_components: struct<feasible_scarcity: double, pareto_frontier_fraction: double, preference_conflict: double, fav (... 58 chars omitted)
child 0, feasible_scarcity: double
child 1, pareto_frontier_fraction: double
child 2, preference_conflict: double
child 3, favorite_block_fraction: double
child 4, pivotal_seat_burden: double
child 8, bank_position: int64
child 9, parameter_sha256: string
child 10, source_sha256: string
n_instances: int64
invariants: struct<public_all_sheets: bool, chat: bool, n_parties: int64, min_accept: int64, integral_thresholds (... 56 chars omitted)
child 0, public_all_sheets: bool
child 1, chat: bool
child 2, n_parties: int64
child 3, min_accept: int64
child 4, integral_thresholds: bool
child 5, economic_parameters_unchanged_from_source: bool
split: string
source_manifest_sha256: string
schema: string
to
{'schema': Value('string'), 'split': Value('string'), 'purpose': Value('string'), 'n_instances': Value('int64'), 'source_bank': Value('string'), 'source_manifest_sha256': Value('string'), 'generation': {'info': Value('string'), 'chat': Value('bool'), 'n_parties': Value('int64'), 'n_issues': Value('int64'), 'n_options': Value('int64'), 'rounds': Value('int64'), 'min_accept': Value('int64'), 'thresholds': Value('string'), 'level_schedule': Value('string'), 'levels': List(Value('int64'))}, 'invariants': {'public_all_sheets': Value('bool'), 'chat': Value('bool'), 'n_parties': Value('int64'), 'min_accept': Value('int64'), 'integral_thresholds': Value('bool'), 'economic_parameters_unchanged_from_source': Value('bool')}, 'instances': List({'instance_id': Value('string'), 'generator_seed': Value('int64'), 'level': Value('int64'), 'sha256': Value('string'), 'content_sha256': Value('string'), 'difficulty_scalar': Value('float64'), 'difficulty_tags': List(Value('string')), 'difficulty_components': {'feasible_scarcity': Value('float64'), 'pareto_frontier_fraction': Value('float64'), 'preference_conflict': Value('float64'), 'favorite_block_fraction': Value('float64'), 'pivotal_seat_burden': Value('float64')}, 'bank_position': Value('int64'), 'parameter_sha256': Value('string'), 'source_sha256': 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
instance_id: string
scenario: string
level: int64
seed: int64
payload: struct<space: struct<issues: list<item: struct<name: string, options: list<item: string>>>>, sheets: (... 823 chars omitted)
child 0, space: struct<issues: list<item: struct<name: string, options: list<item: string>>>>
child 0, issues: list<item: struct<name: string, options: list<item: string>>>
child 0, item: struct<name: string, options: list<item: string>>
child 0, name: string
child 1, options: list<item: string>
child 0, item: string
child 1, sheets: list<item: struct<agent: string, values: list<item: list<item: double>>, threshold: double>>
child 0, item: struct<agent: string, values: list<item: list<item: double>>, threshold: double>
child 0, agent: string
child 1, values: list<item: list<item: double>>
child 0, item: list<item: double>
child 0, item: double
child 2, threshold: double
child 2, rounds: int64
child 3, info: string
child 4, chat: bool
child 5, proposer: int64
child 6, veto: int64
child 7, min_accept: int64
child 8, discount: double
child 9, breakdown_risk: double
child 10, constraint: null
child 11, meta: struct<generator: string, seed: int64, issue_types: list<item: string>, option_perms: list<item: lis (... 471 chars omitted)
child 0, generator: string
child 1, seed: int64
child 2, issue_types: list<item: string>
...
ist<item: struct<instance_id: string, generator_seed: int64, level: int64, sha256: string, content_ (... 334 chars omitted)
child 0, item: struct<instance_id: string, generator_seed: int64, level: int64, sha256: string, content_sha256: str (... 322 chars omitted)
child 0, instance_id: string
child 1, generator_seed: int64
child 2, level: int64
child 3, sha256: string
child 4, content_sha256: string
child 5, difficulty_scalar: double
child 6, difficulty_tags: list<item: string>
child 0, item: string
child 7, difficulty_components: struct<feasible_scarcity: double, pareto_frontier_fraction: double, preference_conflict: double, fav (... 58 chars omitted)
child 0, feasible_scarcity: double
child 1, pareto_frontier_fraction: double
child 2, preference_conflict: double
child 3, favorite_block_fraction: double
child 4, pivotal_seat_burden: double
child 8, bank_position: int64
child 9, parameter_sha256: string
child 10, source_sha256: string
n_instances: int64
invariants: struct<public_all_sheets: bool, chat: bool, n_parties: int64, min_accept: int64, integral_thresholds (... 56 chars omitted)
child 0, public_all_sheets: bool
child 1, chat: bool
child 2, n_parties: int64
child 3, min_accept: int64
child 4, integral_thresholds: bool
child 5, economic_parameters_unchanged_from_source: bool
split: string
source_manifest_sha256: string
schema: string
to
{'schema': Value('string'), 'split': Value('string'), 'purpose': Value('string'), 'n_instances': Value('int64'), 'source_bank': Value('string'), 'source_manifest_sha256': Value('string'), 'generation': {'info': Value('string'), 'chat': Value('bool'), 'n_parties': Value('int64'), 'n_issues': Value('int64'), 'n_options': Value('int64'), 'rounds': Value('int64'), 'min_accept': Value('int64'), 'thresholds': Value('string'), 'level_schedule': Value('string'), 'levels': List(Value('int64'))}, 'invariants': {'public_all_sheets': Value('bool'), 'chat': Value('bool'), 'n_parties': Value('int64'), 'min_accept': Value('int64'), 'integral_thresholds': Value('bool'), 'economic_parameters_unchanged_from_source': Value('bool')}, 'instances': List({'instance_id': Value('string'), 'generator_seed': Value('int64'), 'level': Value('int64'), 'sha256': Value('string'), 'content_sha256': Value('string'), 'difficulty_scalar': Value('float64'), 'difficulty_tags': List(Value('string')), 'difficulty_components': {'feasible_scarcity': Value('float64'), 'pareto_frontier_fraction': Value('float64'), 'preference_conflict': Value('float64'), 'favorite_block_fraction': Value('float64'), 'pivotal_seat_burden': Value('float64')}, 'bank_position': Value('int64'), 'parameter_sha256': Value('string'), 'source_sha256': 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.
PUBLIC Five-Seat Oracle Gap and Prompt-Policy Bandit
This dataset is the complete evidence bundle for a fully public datacenter negotiation experiment. All five
score sheets and whole-number thresholds were common knowledge. Four frozen API prompt policies competed on a
24-game training bank; P0 was locked before evaluation on a disjoint 24-game held-out bank. Canonical
P0, the selected policy, and five exact oracle agents each have 120 matched held-out episodes.
Here, oracle is the repository's exact sequential omniscient best-response policy: it optimizes each turn from
the live state but is not a jointly coordinated global optimizer. A negative oracle gap means the LLM arm
outperformed this oracle policy, not the global feasible optimum.
The primary paired estimands are:
- selected minus baseline: -0.011695833333333332;
- oracle minus selected: -0.13159166666666666;
- oracle-gap closure fraction: None (defined only for a positive initial gap).
See analysis/analysis.md and report/report.pdf for clustered intervals and interpretation. banks/ contains
the hash-verified train and held-out instances. runs/ contains exact instances, raw episodes, API telemetry,
dual turn annotations, Markdown/HTML transcripts, and interactive visualizers. campaign/ records prompt text
and hashes, deterministic rankings, the pre-held-out selection lock, immutable attempts, and the fail-closed
$800 public spend ledger. The program also reserves $1,800 for the Opus private campaign, $180 for discarded
pre-confirmatory attempts, and $190 for robustness checks, keeping the hard envelope at $2,970.
No hidden chain of thought is claimed; raw records retain only provider-returned summarized reasoning. No Weights & Biases runs were used.
Experiment-name mapping
Every packaged CSV row has experiment-name=2026.RA.Public-Oracle-Gap. The arm/role, policy_id, split,
instance id, and seed columns distinguish the four policy-search cells, two finalist cells, and three matched
held-out baseline/selected/oracle cells.
Exact regeneration and upload
cd /workspace/projects/mats-ii-five-seat-20260804/experiments/rational_agents
uv run python launch_public_prompt_bandit.py --train-bank /workspace/projects/mats-ii-five-seat-20260804/experiments/rational_agents/instances_five_seat_public_bandit_v1/train --heldout-bank /workspace/projects/mats-ii-five-seat-20260804/experiments/rational_agents/instances_five_seat_public_bandit_v1/heldout --out /workspace/large_artifacts/ii_mats/rational_agents/control_public_oracle_gap_opus_v3 --campaign-name public_oracle_gap_opus_v3 --artifact-root /workspace/large_artifacts --model anthropic:claude-opus-5 --budget-usd 800.0 --episode-concurrency 4 --api-turn-token-floor 16384 --worst-case-usd-per-request 6.0 --hf-repo siddharthmb/2026.RA.Public-Oracle-Gap --execute --resume
uv run python analyze_public_prompt_bandit.py --manifest /workspace/large_artifacts/ii_mats/rational_agents/control_public_oracle_gap_opus_v3/campaign_manifest.json --out /workspace/large_artifacts/ii_mats/rational_agents/control_public_oracle_gap_opus_v3/analysis
uv run python write_public_prompt_bandit_report.py --manifest /workspace/large_artifacts/ii_mats/rational_agents/control_public_oracle_gap_opus_v3/campaign_manifest.json --analysis /workspace/large_artifacts/ii_mats/rational_agents/control_public_oracle_gap_opus_v3/analysis --out /workspace/large_artifacts/ii_mats/rational_agents/control_public_oracle_gap_opus_v3/report --compile auto --hf-url https://huggingface.co/datasets/siddharthmb/2026.RA.Public-Oracle-Gap
uv run python package_hf_public_prompt_bandit.py --manifest /workspace/large_artifacts/ii_mats/rational_agents/control_public_oracle_gap_opus_v3/campaign_manifest.json --analysis /workspace/large_artifacts/ii_mats/rational_agents/control_public_oracle_gap_opus_v3/analysis --report /workspace/large_artifacts/ii_mats/rational_agents/control_public_oracle_gap_opus_v3/report --out /workspace/large_artifacts/ii_mats/rational_agents/control_public_oracle_gap_opus_v3/hf_bundle
uv run hf repo create siddharthmb/2026.RA.Public-Oracle-Gap --repo-type dataset --exist-ok
uv run hf upload-large-folder siddharthmb/2026.RA.Public-Oracle-Gap /workspace/large_artifacts/ii_mats/rational_agents/control_public_oracle_gap_opus_v3/hf_bundle --repo-type dataset --num-workers 8
Run and artifact locations
- Campaign manifest and orchestration logs:
/workspace/large_artifacts/ii_mats/rational_agents/control_public_oracle_gap_opus_v3/ - Immutable cell logs:
/workspace/large_artifacts/ii_mats/rational_agents/public_oracle_gap_opus_v3__*/run.log - Pod-restart recovery log:
/workspace/large_artifacts/ii_mats/rational_agents/public_resume_after_pod_restart.log - Analysis and report:
/workspace/large_artifacts/ii_mats/rational_agents/control_public_oracle_gap_opus_v3/analysis,/workspace/large_artifacts/ii_mats/rational_agents/control_public_oracle_gap_opus_v3/report
- Downloads last month
- 1,177