Datasets:

yinita
/

ps4mas-tokenhub-tiny-eval-rollouts

Tags:

License:

Dataset card Files Files and versions

xet

Community

Dataset Viewer

Duplicate

Split (1)

train

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
model_name: string
n_traces: int64
n_valid_scores: int64
avg_composite_overall: double
avg_risk_overall: double
avg_empathy_overall: double
avg_personalization_overall: double
mas_mean: double
per_config: struct<PS-cold-single: struct<n: int64, n_invalid: int64, avg_composite: double, avg_risk: double, a (... 749 chars omitted)
  child 0, PS-cold-single: struct<n: int64, n_invalid: int64, avg_composite: double, avg_risk: double, avg_empathy: double, avg (... 93 chars omitted)
      child 0, n: int64
      child 1, n_invalid: int64
      child 2, avg_composite: double
      child 3, avg_risk: double
      child 4, avg_empathy: double
      child 5, avg_personalization: double
      child 6, zero_tool_pct: double
      child 7, mem_first_pct: double
      child 8, any_tool_pct: double
  child 1, PS-cold-central: struct<n: int64, n_invalid: int64, avg_composite: double, avg_risk: double, avg_empathy: double, avg (... 93 chars omitted)
      child 0, n: int64
      child 1, n_invalid: int64
      child 2, avg_composite: double
      child 3, avg_risk: double
      child 4, avg_empathy: double
      child 5, avg_personalization: double
      child 6, zero_tool_pct: double
      child 7, mem_first_pct: double
      child 8, any_tool_pct: double
  child 2, PS-cold-hier: struct<n: int64, n_invalid: int64, avg_composite: double, avg_risk: double, avg_empathy: double, avg (... 93 chars omitted)
      child 0, n: int64
      child 1, n_invalid: int64
      child 2, avg_composite: double
...
e: struct<n: int64, n_invalid: int64, avg_composite: double, avg_risk: double, avg_empathy: double, avg (... 93 chars omitted)
      child 0, n: int64
      child 1, n_invalid: int64
      child 2, avg_composite: double
      child 3, avg_risk: double
      child 4, avg_empathy: double
      child 5, avg_personalization: double
      child 6, zero_tool_pct: double
      child 7, mem_first_pct: double
      child 8, any_tool_pct: double
single_zero_tool_pct: double
single_mem_first_pct: double
recomputed_at: string
sum_agent_output: int64
sum_agent_input: int64
models: list<item: struct<agent_model: string, agent_usage: struct<input_tokens: int64, output_tokens: int64 (... 204 chars omitted)
  child 0, item: struct<agent_model: string, agent_usage: struct<input_tokens: int64, output_tokens: int64, calls: in (... 192 chars omitted)
      child 0, agent_model: string
      child 1, agent_usage: struct<input_tokens: int64, output_tokens: int64, calls: int64>
          child 0, input_tokens: int64
          child 1, output_tokens: int64
          child 2, calls: int64
      child 2, user_sim_usage: struct<input_tokens: int64, output_tokens: int64, calls: int64>
          child 0, input_tokens: int64
          child 1, output_tokens: int64
          child 2, calls: int64
      child 3, n_episodes: int64
      child 4, n_fail: int64
      child 5, total_usd_est: double
      child 6, agent_usd_est: double
      child 7, user_sim_usd_est: double
price_note: string
sum_usd_est: double
to
{'models': List({'agent_model': Value('string'), 'agent_usage': {'input_tokens': Value('int64'), 'output_tokens': Value('int64'), 'calls': Value('int64')}, 'user_sim_usage': {'input_tokens': Value('int64'), 'output_tokens': Value('int64'), 'calls': Value('int64')}, 'n_episodes': Value('int64'), 'n_fail': Value('int64'), 'total_usd_est': Value('float64'), 'agent_usd_est': Value('float64'), 'user_sim_usd_est': Value('float64')}), 'sum_usd_est': Value('float64'), 'sum_agent_input': Value('int64'), 'sum_agent_output': Value('int64'), 'recomputed_at': Value('string'), 'price_note': 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
              model_name: string
              n_traces: int64
              n_valid_scores: int64
              avg_composite_overall: double
              avg_risk_overall: double
              avg_empathy_overall: double
              avg_personalization_overall: double
              mas_mean: double
              per_config: struct<PS-cold-single: struct<n: int64, n_invalid: int64, avg_composite: double, avg_risk: double, a (... 749 chars omitted)
                child 0, PS-cold-single: struct<n: int64, n_invalid: int64, avg_composite: double, avg_risk: double, avg_empathy: double, avg (... 93 chars omitted)
                    child 0, n: int64
                    child 1, n_invalid: int64
                    child 2, avg_composite: double
                    child 3, avg_risk: double
                    child 4, avg_empathy: double
                    child 5, avg_personalization: double
                    child 6, zero_tool_pct: double
                    child 7, mem_first_pct: double
                    child 8, any_tool_pct: double
                child 1, PS-cold-central: struct<n: int64, n_invalid: int64, avg_composite: double, avg_risk: double, avg_empathy: double, avg (... 93 chars omitted)
                    child 0, n: int64
                    child 1, n_invalid: int64
                    child 2, avg_composite: double
                    child 3, avg_risk: double
                    child 4, avg_empathy: double
                    child 5, avg_personalization: double
                    child 6, zero_tool_pct: double
                    child 7, mem_first_pct: double
                    child 8, any_tool_pct: double
                child 2, PS-cold-hier: struct<n: int64, n_invalid: int64, avg_composite: double, avg_risk: double, avg_empathy: double, avg (... 93 chars omitted)
                    child 0, n: int64
                    child 1, n_invalid: int64
                    child 2, avg_composite: double
              ...
              e: struct<n: int64, n_invalid: int64, avg_composite: double, avg_risk: double, avg_empathy: double, avg (... 93 chars omitted)
                    child 0, n: int64
                    child 1, n_invalid: int64
                    child 2, avg_composite: double
                    child 3, avg_risk: double
                    child 4, avg_empathy: double
                    child 5, avg_personalization: double
                    child 6, zero_tool_pct: double
                    child 7, mem_first_pct: double
                    child 8, any_tool_pct: double
              single_zero_tool_pct: double
              single_mem_first_pct: double
              recomputed_at: string
              sum_agent_output: int64
              sum_agent_input: int64
              models: list<item: struct<agent_model: string, agent_usage: struct<input_tokens: int64, output_tokens: int64 (... 204 chars omitted)
                child 0, item: struct<agent_model: string, agent_usage: struct<input_tokens: int64, output_tokens: int64, calls: in (... 192 chars omitted)
                    child 0, agent_model: string
                    child 1, agent_usage: struct<input_tokens: int64, output_tokens: int64, calls: int64>
                        child 0, input_tokens: int64
                        child 1, output_tokens: int64
                        child 2, calls: int64
                    child 2, user_sim_usage: struct<input_tokens: int64, output_tokens: int64, calls: int64>
                        child 0, input_tokens: int64
                        child 1, output_tokens: int64
                        child 2, calls: int64
                    child 3, n_episodes: int64
                    child 4, n_fail: int64
                    child 5, total_usd_est: double
                    child 6, agent_usd_est: double
                    child 7, user_sim_usd_est: double
              price_note: string
              sum_usd_est: double
              to
              {'models': List({'agent_model': Value('string'), 'agent_usage': {'input_tokens': Value('int64'), 'output_tokens': Value('int64'), 'calls': Value('int64')}, 'user_sim_usage': {'input_tokens': Value('int64'), 'output_tokens': Value('int64'), 'calls': Value('int64')}, 'n_episodes': Value('int64'), 'n_fail': Value('int64'), 'total_usd_est': Value('float64'), 'agent_usd_est': Value('float64'), 'user_sim_usd_est': Value('float64')}), 'sum_usd_est': Value('float64'), 'sum_agent_input': Value('int64'), 'sum_agent_output': Value('int64'), 'recomputed_at': Value('string'), 'price_note': 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.

PS4MAS TokenHub tiny_eval rollouts (cold×4)

Inference-only trajectories: 50 scenarios × 4 cold topologies × 4 agents. Judge (OSS-120B) deferred.

Agents: claude-opus-5, gpt-5.4, qwen3.7-plus, deepseek-v4-pro via TokenHub. Protocol: B=3, T=0, max_steps=6, seed=0.

See protocol.json and per-model usage.json / cost_summary.json.

OSS-120B judge results (added)

Sealed-protocol rejudge of all rollouts with openai/gpt-oss-120b (same prompt as paper tiny_eval).

Model	Overall	Single	Central	Hier	Debate	MAS	Single zero	Single mem1
deepseek-v4-pro	4.014	3.654	4.013	4.014	4.374	4.134	8%	74%
claude-opus-5	3.971	3.948	3.840	4.194	3.900	3.978	4%	96%
qwen3.7-plus	3.403	2.726	3.460	3.387	4.041	3.629	50%	32%
gpt-5.4	2.726	2.553	2.814	2.513	3.026	2.784	70%	30%

Paths:

oss_judge/RESULTS.md
oss_judge/<model>/traces.jsonl (rollouts + scores)
oss_judge/<model>/summary.json
oss_judge/all_summaries.json

Downloads last month: 429

Total file size:

102 MB