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
model: string
benchmark: string
thinking: bool
runs: int64
tasks_per_run: int64
run_success: struct<1: string, 2: string>
child 0, 1: string
child 1, 2: string
mean_success_rate: double
combined: string
pass_at_2: int64
time: double
agent: struct<type: string, class: string, name: string, prompt_name: string, screenshot_only: bool, histor (... 98 chars omitted)
child 0, type: string
child 1, class: string
child 2, name: string
child 3, prompt_name: string
child 4, screenshot_only: bool
child 5, history_len: int64
child 6, n_history_image: int64
child 7, wait_after_action_seconds: double
child 8, hint: bool
child 9, note: string
served_model_name: string
endpoint: string
evaluation: struct<suite_family: string, tasks: int64, workers: int64, seeds: list<item: int64>, repeats_per_see (... 151 chars omitted)
child 0, suite_family: string
child 1, tasks: int64
child 2, workers: int64
child 3, seeds: list<item: int64>
child 0, item: int64
child 4, repeats_per_seed: int64
child 5, seed_semantics: string
child 6, step_budget: string
child 7, entry: string
child 8, config_name: string
child 9, docker_image_built: string
child 10, avd: string
child 11, avd_volume: string
eval_tag: string
sampling: struct<temperature: int64, top_p: int64, max_tokens: int64, request_seed: null, thinking: bool, engi (... 29 chars omitted)
child 0, temperature: int64
child 1, top_p: int64
child 2, max_tokens: int64
child 3, request_seed: null
child 4, thinking: bool
child 5, engine_seed: int64
child 6, note: string
campaign: string
serving: struct<vllm: string, instances: list<item: struct<gpus: list<item: int64>, port: int64, tensor_paral (... 147 chars omitted)
child 0, vllm: string
child 1, instances: list<item: struct<gpus: list<item: int64>, port: int64, tensor_parallel: int64, gpu_memory_utilizati (... 114 chars omitted)
child 0, item: struct<gpus: list<item: int64>, port: int64, tensor_parallel: int64, gpu_memory_utilization: double, (... 102 chars omitted)
child 0, gpus: list<item: int64>
child 0, item: int64
child 1, port: int64
child 2, tensor_parallel: int64
child 3, gpu_memory_utilization: double
child 4, max_model_len: int64
child 5, max_num_seqs: int64
child 6, seed: int64
child 7, extra_args: list<item: string>
child 0, item: string
child 8, note: string
to
{'model': Value('string'), 'served_model_name': Value('string'), 'thinking': Value('bool'), 'serving': {'vllm': Value('string'), 'instances': List({'gpus': List(Value('int64')), 'port': Value('int64'), 'tensor_parallel': Value('int64'), 'gpu_memory_utilization': Value('float64'), 'max_model_len': Value('int64'), 'max_num_seqs': Value('int64'), 'seed': Value('int64'), 'extra_args': List(Value('string')), 'note': Value('string')})}, 'agent': {'type': Value('string'), 'class': Value('string'), 'name': Value('string'), 'prompt_name': Value('string'), 'screenshot_only': Value('bool'), 'history_len': Value('int64'), 'n_history_image': Value('int64'), 'wait_after_action_seconds': Value('float64'), 'hint': Value('bool'), 'note': Value('string')}, 'sampling': {'temperature': Value('int64'), 'top_p': Value('int64'), 'max_tokens': Value('int64'), 'request_seed': Value('null'), 'thinking': Value('bool'), 'engine_seed': Value('int64'), 'note': Value('string')}, 'evaluation': {'suite_family': Value('string'), 'tasks': Value('int64'), 'workers': Value('int64'), 'seeds': List(Value('int64')), 'repeats_per_seed': Value('int64'), 'seed_semantics': Value('string'), 'step_budget': Value('string'), 'entry': Value('string'), 'config_name': Value('string'), 'docker_image_built': Value('string'), 'avd': Value('string'), 'avd_volume': Value('string')}, 'endpoint': Value('string'), 'campaign': Value('string'), 'eval_tag': Value('string'), 'time': Value('float64')}
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 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 361, 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 2381, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
model: string
benchmark: string
thinking: bool
runs: int64
tasks_per_run: int64
run_success: struct<1: string, 2: string>
child 0, 1: string
child 1, 2: string
mean_success_rate: double
combined: string
pass_at_2: int64
time: double
agent: struct<type: string, class: string, name: string, prompt_name: string, screenshot_only: bool, histor (... 98 chars omitted)
child 0, type: string
child 1, class: string
child 2, name: string
child 3, prompt_name: string
child 4, screenshot_only: bool
child 5, history_len: int64
child 6, n_history_image: int64
child 7, wait_after_action_seconds: double
child 8, hint: bool
child 9, note: string
served_model_name: string
endpoint: string
evaluation: struct<suite_family: string, tasks: int64, workers: int64, seeds: list<item: int64>, repeats_per_see (... 151 chars omitted)
child 0, suite_family: string
child 1, tasks: int64
child 2, workers: int64
child 3, seeds: list<item: int64>
child 0, item: int64
child 4, repeats_per_seed: int64
child 5, seed_semantics: string
child 6, step_budget: string
child 7, entry: string
child 8, config_name: string
child 9, docker_image_built: string
child 10, avd: string
child 11, avd_volume: string
eval_tag: string
sampling: struct<temperature: int64, top_p: int64, max_tokens: int64, request_seed: null, thinking: bool, engi (... 29 chars omitted)
child 0, temperature: int64
child 1, top_p: int64
child 2, max_tokens: int64
child 3, request_seed: null
child 4, thinking: bool
child 5, engine_seed: int64
child 6, note: string
campaign: string
serving: struct<vllm: string, instances: list<item: struct<gpus: list<item: int64>, port: int64, tensor_paral (... 147 chars omitted)
child 0, vllm: string
child 1, instances: list<item: struct<gpus: list<item: int64>, port: int64, tensor_parallel: int64, gpu_memory_utilizati (... 114 chars omitted)
child 0, item: struct<gpus: list<item: int64>, port: int64, tensor_parallel: int64, gpu_memory_utilization: double, (... 102 chars omitted)
child 0, gpus: list<item: int64>
child 0, item: int64
child 1, port: int64
child 2, tensor_parallel: int64
child 3, gpu_memory_utilization: double
child 4, max_model_len: int64
child 5, max_num_seqs: int64
child 6, seed: int64
child 7, extra_args: list<item: string>
child 0, item: string
child 8, note: string
to
{'model': Value('string'), 'served_model_name': Value('string'), 'thinking': Value('bool'), 'serving': {'vllm': Value('string'), 'instances': List({'gpus': List(Value('int64')), 'port': Value('int64'), 'tensor_parallel': Value('int64'), 'gpu_memory_utilization': Value('float64'), 'max_model_len': Value('int64'), 'max_num_seqs': Value('int64'), 'seed': Value('int64'), 'extra_args': List(Value('string')), 'note': Value('string')})}, 'agent': {'type': Value('string'), 'class': Value('string'), 'name': Value('string'), 'prompt_name': Value('string'), 'screenshot_only': Value('bool'), 'history_len': Value('int64'), 'n_history_image': Value('int64'), 'wait_after_action_seconds': Value('float64'), 'hint': Value('bool'), 'note': Value('string')}, 'sampling': {'temperature': Value('int64'), 'top_p': Value('int64'), 'max_tokens': Value('int64'), 'request_seed': Value('null'), 'thinking': Value('bool'), 'engine_seed': Value('int64'), 'note': Value('string')}, 'evaluation': {'suite_family': Value('string'), 'tasks': Value('int64'), 'workers': Value('int64'), 'seeds': List(Value('int64')), 'repeats_per_seed': Value('int64'), 'seed_semantics': Value('string'), 'step_budget': Value('string'), 'entry': Value('string'), 'config_name': Value('string'), 'docker_image_built': Value('string'), 'avd': Value('string'), 'avd_volume': Value('string')}, 'endpoint': Value('string'), 'campaign': Value('string'), 'eval_tag': Value('string'), 'time': Value('float64')}
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.
Qwen3.8-27B — AndroidWorld — THINKING MODE
Two fresh runs of the native qwen agent (QwenAgent, prompt qwen3vl_instruct, no hint
augmentation) against AndroidWorld (116 task templates) with task seeds 1 and 2 and
thinking mode enabled. Seeds change task parameters (names, dates, amounts, instruction
wording) and initial state.
Serving: local vLLM 0.28.0, Qwen3.8-27B bf16, TP=2 (GPU 0,1), greedy sampling
(temperature 0, top_p 1, max_tokens 8192, enable_thinking: true both server-side default
and per-request extra_body), official default step budgets. No parser patch needed: the
agent extracts <tool_call> from the whole response; the full thinking text is preserved
verbatim in every step record.
Non-thinking baseline: GUI-Gym/qwen3.8-27b-android-world (56/116 and 64/116, mean 51.72%).
| Run | Seed | Success | Unscored |
|---|---|---|---|
| 1 | 1 | 63/116 | 0 |
| 2 | 2 | 71/116 | 0 |
Mean success rate: 134/232 = 57.76%. Pass@2 (task templates solved in at least one run): 80/116.
runs/run_N.jsonl— one row per task attempt (success, steps, budget, trajectory path).runs/run_N_trajectories.tar.gz— native per-task records: step-wise conversations (repeat_0X_{succ,fail}.jsonl, with full thinking text) and screenshots (images/).two_runs/— combined episodes, protocol, summary, package manifest.
- Downloads last month
- 50