Datasets:

LisaMegaWatts
/

sshfighter-current-epoch-baseline-459

Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
schema_version: int64
match_id: string
base_seed: int64
scenario_seed: int64
policy: string
policy_a: string
policy_b: string
controller: string
controller_a: string
controller_b: string
fighter_a: string
fighter_b: string
paired_key: string
mirror: bool
character_mirror: bool
policy_mirror: bool
controller_mirror: bool
stage: string
completed: bool
match_status: string
result_a: string
winner: string
end_reason: string
failure_kind: null
efficacy_valid: bool
rounds_a: int64
rounds_b: int64
frames: int64
detector_policy_order: list<item: string>
  child 0, item: string
detector_action_classes: list<item: string>
  child 0, item: string
detector_timeline_encoding: string
detector_timeline_b64: string
detector_class_counts: struct<a: struct<champion: list<item: int64>, grappler: list<item: int64>, hitrun: list<item: int64> (... 422 chars omitted)
  child 0, a: struct<champion: list<item: int64>, grappler: list<item: int64>, hitrun: list<item: int64>, jumper:  (... 153 chars omitted)
      child 0, champion: list<item: int64>
          child 0, item: int64
      child 1, grappler: list<item: int64>
          child 0, item: int64
      child 2, hitrun: list<item: int64>
          child 0, item: int64
      child 3, jumper: list<item: int64>
          child 0, item: int64
      child 4, masher: list<item: int64>
          child 0, item: int64
      child 5, poker: list<item: int64>
          child 0, item: int64
      child 6, rushdown: list<item: int64>
          child 0, item: i
...
ist<item: string>
  child 0, item: string
stalemates: int64
active_policies: list<item: string>
  child 0, item: string
source_profile: string
limitations: list<item: string>
  child 0, item: string
expected_matches: int64
ssh_street_fighter_commit: string
seeds: list<item: int64>
  child 0, item: int64
started_at: string
network_access: bool
source_hashes: struct<engine: string, types: string, moves: string, roster: string, gym: string, coordinator: strin (... 37 chars omitted)
  child 0, engine: string
  child 1, types: string
  child 2, moves: string
  child 3, roster: string
  child 4, gym: string
  child 5, coordinator: string
  child 6, messages: string
  child 7, harness: string
passive_detector_policies: list<item: string>
  child 0, item: string
pair_units: int64
stage_control: string
selection_or_tuning: bool
balance_intervention_treatment: string
aggregate: struct<aWins: int64, bWins: int64, draws: int64, stalemates: int64, frames: int64>
  child 0, aWins: int64
  child 1, bWins: int64
  child 2, draws: int64
  child 3, stalemates: int64
  child 4, frames: int64
offline_only: bool
timeline: struct<policy_order: list<item: string>, action_classes: list<item: string>, bytes_per_frame: int64, (... 14 chars omitted)
  child 0, policy_order: list<item: string>
      child 0, item: string
  child 1, action_classes: list<item: string>
      child 0, item: string
  child 2, bytes_per_frame: int64
  child 3, note: string
written_matches: int64
status: string
failures: int64
to
{'schema_version': Value('int64'), 'experiment': Value('string'), 'design': Value('string'), 'status': Value('string'), 'started_at': Value('string'), 'completed_at': Value('string'), 'offline_only': Value('bool'), 'network_access': Value('bool'), 'selection_or_tuning': Value('bool'), 'source_profile': Value('string'), 'experimental_epoch': Value('string'), 'balance_intervention_treatment': Value('string'), 'stage_control': Value('string'), 'ssh_street_fighter_commit': Value('string'), 'source_hashes': {'engine': Value('string'), 'types': Value('string'), 'moves': Value('string'), 'roster': Value('string'), 'gym': Value('string'), 'coordinator': Value('string'), 'messages': Value('string'), 'harness': Value('string')}, 'roster': List(Value('string')), 'active_policies': List(Value('string')), 'controllers': List(Value('string')), 'passive_detector_policies': List(Value('string')), 'seeds': List(Value('int64')), 'pair_units': Value('int64'), 'expected_matches': Value('int64'), 'written_matches': Value('int64'), 'failures': Value('int64'), 'stalemates': Value('int64'), 'aggregate': {'aWins': Value('int64'), 'bWins': Value('int64'), 'draws': Value('int64'), 'stalemates': Value('int64'), 'frames': Value('int64')}, 'timeline': {'policy_order': List(Value('string')), 'action_classes': List(Value('string')), 'bytes_per_frame': Value('int64'), 'note': Value('string')}, 'limitations': List(Value('string'))}
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 483, 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 2840, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, 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 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              schema_version: int64
              match_id: string
              base_seed: int64
              scenario_seed: int64
              policy: string
              policy_a: string
              policy_b: string
              controller: string
              controller_a: string
              controller_b: string
              fighter_a: string
              fighter_b: string
              paired_key: string
              mirror: bool
              character_mirror: bool
              policy_mirror: bool
              controller_mirror: bool
              stage: string
              completed: bool
              match_status: string
              result_a: string
              winner: string
              end_reason: string
              failure_kind: null
              efficacy_valid: bool
              rounds_a: int64
              rounds_b: int64
              frames: int64
              detector_policy_order: list<item: string>
                child 0, item: string
              detector_action_classes: list<item: string>
                child 0, item: string
              detector_timeline_encoding: string
              detector_timeline_b64: string
              detector_class_counts: struct<a: struct<champion: list<item: int64>, grappler: list<item: int64>, hitrun: list<item: int64> (... 422 chars omitted)
                child 0, a: struct<champion: list<item: int64>, grappler: list<item: int64>, hitrun: list<item: int64>, jumper:  (... 153 chars omitted)
                    child 0, champion: list<item: int64>
                        child 0, item: int64
                    child 1, grappler: list<item: int64>
                        child 0, item: int64
                    child 2, hitrun: list<item: int64>
                        child 0, item: int64
                    child 3, jumper: list<item: int64>
                        child 0, item: int64
                    child 4, masher: list<item: int64>
                        child 0, item: int64
                    child 5, poker: list<item: int64>
                        child 0, item: int64
                    child 6, rushdown: list<item: int64>
                        child 0, item: i
              ...
              ist<item: string>
                child 0, item: string
              stalemates: int64
              active_policies: list<item: string>
                child 0, item: string
              source_profile: string
              limitations: list<item: string>
                child 0, item: string
              expected_matches: int64
              ssh_street_fighter_commit: string
              seeds: list<item: int64>
                child 0, item: int64
              started_at: string
              network_access: bool
              source_hashes: struct<engine: string, types: string, moves: string, roster: string, gym: string, coordinator: strin (... 37 chars omitted)
                child 0, engine: string
                child 1, types: string
                child 2, moves: string
                child 3, roster: string
                child 4, gym: string
                child 5, coordinator: string
                child 6, messages: string
                child 7, harness: string
              passive_detector_policies: list<item: string>
                child 0, item: string
              pair_units: int64
              stage_control: string
              selection_or_tuning: bool
              balance_intervention_treatment: string
              aggregate: struct<aWins: int64, bWins: int64, draws: int64, stalemates: int64, frames: int64>
                child 0, aWins: int64
                child 1, bWins: int64
                child 2, draws: int64
                child 3, stalemates: int64
                child 4, frames: int64
              offline_only: bool
              timeline: struct<policy_order: list<item: string>, action_classes: list<item: string>, bytes_per_frame: int64, (... 14 chars omitted)
                child 0, policy_order: list<item: string>
                    child 0, item: string
                child 1, action_classes: list<item: string>
                    child 0, item: string
                child 2, bytes_per_frame: int64
                child 3, note: string
              written_matches: int64
              status: string
              failures: int64
              to
              {'schema_version': Value('int64'), 'experiment': Value('string'), 'design': Value('string'), 'status': Value('string'), 'started_at': Value('string'), 'completed_at': Value('string'), 'offline_only': Value('bool'), 'network_access': Value('bool'), 'selection_or_tuning': Value('bool'), 'source_profile': Value('string'), 'experimental_epoch': Value('string'), 'balance_intervention_treatment': Value('string'), 'stage_control': Value('string'), 'ssh_street_fighter_commit': Value('string'), 'source_hashes': {'engine': Value('string'), 'types': Value('string'), 'moves': Value('string'), 'roster': Value('string'), 'gym': Value('string'), 'coordinator': Value('string'), 'messages': Value('string'), 'harness': Value('string')}, 'roster': List(Value('string')), 'active_policies': List(Value('string')), 'controllers': List(Value('string')), 'passive_detector_policies': List(Value('string')), 'seeds': List(Value('int64')), 'pair_units': Value('int64'), 'expected_matches': Value('int64'), 'written_matches': Value('int64'), 'failures': Value('int64'), 'stalemates': Value('int64'), 'aggregate': {'aWins': Value('int64'), 'bWins': Value('int64'), 'draws': Value('int64'), 'stalemates': Value('int64'), 'frames': Value('int64')}, 'timeline': {'policy_order': List(Value('string')), 'action_classes': List(Value('string')), 'bytes_per_frame': Value('int64'), 'note': Value('string')}, 'limitations': List(Value('string'))}
              because column names don't match

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.

SSH Fighter current-epoch matched baseline (459 matches)

Exact-current-main (3caedf3435c12996cf4d34fb5ac76c7cd7b75076) offline baseline: 17 characters × 9 shared Gym policies × 3 fresh seeds. Character, active policy, and direct-fresh controller are held equal on both sides. All nine policies additionally run as passive action-fire detectors and never influence the active match.

The matrix completed 459/459 simulations with zero technical failures: 316 decisive KOs and 143 declared 6,300-frame behavioral stalemates. The 266 A wins versus 50 B wins are role-conditioned observations, not a clean seat causal estimate: the harness deliberately gives the two roles distinct deterministic policy RNG streams, and the engine updates A before B.

The frozen three-seed HP-kick oscillator from the prior Leg-2 data was applied without retraining to 21,536 causal current-epoch shadow samples:

oscillator ensemble: AUROC 0.83594 / AUPRC 0.49181
aggregate policy attack-fire fraction: AUROC 0.58455 / AUPRC 0.35145
fixed descriptive trigger overlap future-hit rates:
- neither: 5.51%
- policy-majority only: 17.84%
- oscillator only: 37.84%
- both: 52.61%

These remain separate signal families. The thresholds are descriptive and were not optimized into a routing heuristic. This is direct exact-engine evidence, not SSH transport or live-match efficacy evidence.

Files:

matches.jsonl: authoritative terminal rows and compact per-frame policy detector timelines
monitor_events.jsonl: passive disagreement, CUSUM, oscillator-proxy, and exact-engine HP-change events
manifest.json: exact source/profile/harness hashes
shadow_results.json: frozen oscillator, policy-fire, overlap, character, and seat summaries
failure_traces.jsonl: empty because there were zero technical failures

Downloads last month: 46

Total file size:

25.8 MB