Dataset Preview
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed because of a cast error
Error code: DatasetGenerationCastError
Exception: DatasetGenerationCastError
Message: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 1 new columns ({'reason'}) and 5 missing columns ({'explanation_q2', 'explanation_q1', 'robotic_phrase', 'object_phrase', 'option'}).
This happened while the csv dataset builder was generating data using
hf://datasets/gyHit/WEAVE-0.5/evaluation/gpt/4_embodiments/dual_arm/VQA/2_physical_plausibility/results.csv (at revision 9270517d5522e70704c16038cd055a67da3a55e3), ['hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/dual_arm/VQA/1_robot_subject_stability/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/dual_arm/VQA/2_physical_plausibility/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/dual_arm/VQA/3_task_adherence_consistency/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/dual_arm/score_summary_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/dual_arm/score_summary_normalized_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/dual_arm/score_summary_penalized_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/humanoid/VQA/1_robot_subject_stability/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/humanoid/VQA/2_physical_plausibility/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/humanoid/VQA/3_task_adherence_consistency/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/humanoid/score_summary_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/humanoid/score_summary_normalized_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/humanoid/score_summary_penalized_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/quad/VQA/1_robot_subject_stability/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/quad/VQA/2_physical_plausibility/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/quad/VQA/3_task_adherence_consistency/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/quad/score_summary_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/quad/score_summary_normalized_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/quad/score_summary_penalized_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/single_arm/VQA/1_robot_subject_stability/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/single_arm/VQA/2_physical_plausibility/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/single_arm/VQA/3_task_adherence_consistency/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/single_arm/score_summary_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/single_arm/score_summary_normalized_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/single_arm/score_summary_penalized_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/5_tasks/common_manipulation/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/5_tasks/long-horizon_planning/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/5_tasks/multi-entity_collaboration/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/5_tasks/spatial_relationship/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/5_tasks/visual_reasoning/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/detailed_results_650.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/score_summary.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/dual_arm/VQA/qwen_api/1_robot_subject_stability/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/dual_arm/VQA/qwen_api/2_physical_plausibility/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/dual_arm/VQA/qwen_api/3_task_adherence_consistency/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/dual_arm/score_summary_normalized_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/dual_arm/score_summary_penalized_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/dual_arm/score_summary_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/humanoid/VQA/qwen_api/1_robot_subject_stability/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/humanoid/VQA/qwen_api/2_physical_plausibility/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/humanoid/VQA/qwen_api/3_task_adherence_consistency/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/humanoid/score_summary_normalized_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/humanoid/score_summary_penalized_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/humanoid/score_summary_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/quad/VQA/qwen_api/1_robot_subject_stability/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/quad/VQA/qwen_api/2_physical_plausibility/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/quad/VQA/qwen_api/3_task_adherence_consistency/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/quad/score_summary_normalized_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/quad/score_summary_penalized_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/quad/score_summary_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/single_arm/VQA/qwen_api/1_robot_subject_stability/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/single_arm/VQA/qwen_api/2_physical_plausibility/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/single_arm/VQA/qwen_api/3_task_adherence_consistency/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/single_arm/score_summary_normalized_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/single_arm/score_summary_penalized_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/single_arm/score_summary_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/5_tasks/rbench_650_qwen_retest2_20260902_qwen235_api/common_manipulation/qwen_api/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/5_tasks/rbench_650_qwen_retest2_20260902_qwen235_api/long-horizon_planning/qwen_api/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/5_tasks/rbench_650_qwen_retest2_20260902_qwen235_api/multi-entity_collaboration/qwen_api/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/5_tasks/rbench_650_qwen_retest2_20260902_qwen235_api/spatial_relationship/qwen_api/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/5_tasks/rbench_650_qwen_retest2_20260902_qwen235_api/visual_reasoning/qwen_api/results.csv']
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1848, in _prepare_split_single
writer.write_table(table)
~~~~~~~~~~~~~~~~~~^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 765, in write_table
self._write_table(pa_table, writer_batch_size=writer_batch_size)
~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 773, in _write_table
pa_table = table_cast(pa_table, self._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
name: string
score: int64
prompt: string
reason: string
-- schema metadata --
pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 718
to
{'name': Value('string'), 'prompt': Value('string'), 'robotic_phrase': Value('string'), 'object_phrase': Value('string'), 'option': Value('string'), 'score': Value('int64'), 'explanation_q1': Value('string'), 'explanation_q2': Value('string')}
because column names don't match
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
~~~~~~~~~~~~~~~~~~~~~~~~~^
builder, max_dataset_size_bytes=max_dataset_size_bytes
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
for job_id, done, content in self._prepare_split_single(
~~~~~~~~~~~~~~~~~~~~~~~~~~^
gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
):
^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1850, in _prepare_split_single
raise DatasetGenerationCastError.from_cast_error(
...<4 lines>...
)
datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 1 new columns ({'reason'}) and 5 missing columns ({'explanation_q2', 'explanation_q1', 'robotic_phrase', 'object_phrase', 'option'}).
This happened while the csv dataset builder was generating data using
hf://datasets/gyHit/WEAVE-0.5/evaluation/gpt/4_embodiments/dual_arm/VQA/2_physical_plausibility/results.csv (at revision 9270517d5522e70704c16038cd055a67da3a55e3), ['hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/dual_arm/VQA/1_robot_subject_stability/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/dual_arm/VQA/2_physical_plausibility/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/dual_arm/VQA/3_task_adherence_consistency/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/dual_arm/score_summary_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/dual_arm/score_summary_normalized_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/dual_arm/score_summary_penalized_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/humanoid/VQA/1_robot_subject_stability/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/humanoid/VQA/2_physical_plausibility/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/humanoid/VQA/3_task_adherence_consistency/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/humanoid/score_summary_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/humanoid/score_summary_normalized_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/humanoid/score_summary_penalized_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/quad/VQA/1_robot_subject_stability/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/quad/VQA/2_physical_plausibility/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/quad/VQA/3_task_adherence_consistency/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/quad/score_summary_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/quad/score_summary_normalized_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/quad/score_summary_penalized_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/single_arm/VQA/1_robot_subject_stability/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/single_arm/VQA/2_physical_plausibility/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/single_arm/VQA/3_task_adherence_consistency/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/single_arm/score_summary_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/single_arm/score_summary_normalized_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/4_embodiments/single_arm/score_summary_penalized_gpt5.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/5_tasks/common_manipulation/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/5_tasks/long-horizon_planning/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/5_tasks/multi-entity_collaboration/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/5_tasks/spatial_relationship/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/5_tasks/visual_reasoning/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/detailed_results_650.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/gpt/score_summary.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/dual_arm/VQA/qwen_api/1_robot_subject_stability/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/dual_arm/VQA/qwen_api/2_physical_plausibility/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/dual_arm/VQA/qwen_api/3_task_adherence_consistency/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/dual_arm/score_summary_normalized_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/dual_arm/score_summary_penalized_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/dual_arm/score_summary_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/humanoid/VQA/qwen_api/1_robot_subject_stability/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/humanoid/VQA/qwen_api/2_physical_plausibility/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/humanoid/VQA/qwen_api/3_task_adherence_consistency/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/humanoid/score_summary_normalized_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/humanoid/score_summary_penalized_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/humanoid/score_summary_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/quad/VQA/qwen_api/1_robot_subject_stability/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/quad/VQA/qwen_api/2_physical_plausibility/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/quad/VQA/qwen_api/3_task_adherence_consistency/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/quad/score_summary_normalized_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/quad/score_summary_penalized_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/quad/score_summary_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/single_arm/VQA/qwen_api/1_robot_subject_stability/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/single_arm/VQA/qwen_api/2_physical_plausibility/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/single_arm/VQA/qwen_api/3_task_adherence_consistency/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/single_arm/score_summary_normalized_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/single_arm/score_summary_penalized_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/4_embodiments/rbench_650_qwen_retest2_20260902_qwen235_api/single_arm/score_summary_qwen.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/5_tasks/rbench_650_qwen_retest2_20260902_qwen235_api/common_manipulation/qwen_api/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/5_tasks/rbench_650_qwen_retest2_20260902_qwen235_api/long-horizon_planning/qwen_api/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/5_tasks/rbench_650_qwen_retest2_20260902_qwen235_api/multi-entity_collaboration/qwen_api/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/5_tasks/rbench_650_qwen_retest2_20260902_qwen235_api/spatial_relationship/qwen_api/results.csv', 'hf://datasets/gyHit/WEAVE-0.5@9270517d5522e70704c16038cd055a67da3a55e3/evaluation/qwen/5_tasks/rbench_650_qwen_retest2_20260902_qwen235_api/visual_reasoning/qwen_api/results.csv']
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)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.
name string | prompt string | robotic_phrase string | object_phrase string | option string | score int64 | explanation_q1 string | explanation_q2 string |
|---|---|---|---|---|---|---|---|
dual_arm_0001 | The robotic gripper adjust the pillow on the bed. | robotic grippers | pillow | B1,B2 | 13 | Both frames show the same two three-pronged robotic grippers with matching housings. The right frame has slight blur and small variations in prong angles/lengths due to rendering, but no structural change or hand replacement. Mostly consistent with minor visual issues. | The long beige patterned pillow is present with similar shape and placement. Minor differences in color tone and texture detail likely from lighting or generation artifacts, but it remains clearly recognizable as the same pillow. |
dual_arm_0002 | The robotic gripper pour the coffee from the kettle into the transparent glass. | robotic grippers | kettle | B1,B2 | 13 | Both frames show the same robotic grippers: a grey parallel-jaw gripper near the green joint and a black two-jaw gripper on the right. Their structures match; differences are limited to angle/lighting and slight detail smoothing in the generated frame. No switch to a hand. | The kettle appears in both frames at the right edge with similar shape and position. It’s slightly cropped and less distinct in the generated frame, showing minor blur/definition issues but no major changes or duplication. |
dual_arm_0003 | The robotic gripper place the green bottle inside the gray container. | robotic grippers | green bottle | A1,B2 | 14 | Both frames show the same two‑jaw robotic grippers with consistent structure and materials. The right frame only changes orientation and has slight occlusion while grasping the bottle; no structural changes or switch to a robotic hand are observed. | The green bottle matches in color, shape, and size. In the generated frame it’s being grasped and partly occluded, with slight blur/texture differences. These minor visual issues don’t significantly alter the object, so it’s mostly consistent with the reference. |
dual_arm_0004 | The robotic gripper use a stainless steel spoon to scoop cereal from the cereal jar into the bowl. | robotic grippers | stainless spoon | A1,B2 | 14 | Both frames show the same black three‑jaw robotic grippers with identical structure and housing; only orientation and task differ. The right gripper holds a spoon but remains the same gripper type. No noticeable shape or structural changes. | The stainless spoon appears in both frames, relocated from the bowl to the robot’s gripper. It remains metallic and similar in shape, though reflections and slight shape/length differences suggest minor visual artifacts rather than major changes. |
dual_arm_0005 | The robotic gripper used a jam squeeze bottle to dispense jam onto the bread. | robotic grippers | jam squeeze bottle | B1,C2 | 11 | Both frames show similar two-jaw robotic grippers with consistent structure and finger count. Orientation and slight visual artifacts differ, but no change to a hand or different gripper type is seen. | The jam squeeze bottle appears in both frames with similar red body and white flip cap, but in the generated frame the cap looks duplicated/misaligned and the bottle geometry is distorted, indicating noticeable inconsistencies rather than just minor visual noise. |
dual_arm_0006 | The robotic gripper position the wooden tray on the table mat. | robotic grippers | wooden tray | A1,B2 | 14 | Both frames show the same two-jaw black robotic grippers. The structure and type remain unchanged; only orientation and slight motion blur differ. No transformation into a different gripper or a robotic/human hand is observed. | The wooden tray appears in both frames with matching shape, color, and contents. Minor visual differences (lighting, slight perspective/size shift and slight blur) are present but do not significantly alter its identity. |
dual_arm_0007 | The robotic gripper places the wooden tray on the wooden table. | robotic grippers | wooden tray | B1,B2 | 13 | Both frames show the same parallel-jaw robotic grippers with two rigid prongs on each side. Color, mounting, and general structure match. Minor differences appear due to angle/lighting and small rendering artifacts, but no change to type or function, and not replaced by a hand. | The wooden tray matches in shape, color, and raised rim, remaining clearly identifiable. Minor differences due to perspective/lighting and slight edge or texture artifacts are present, but nothing major or transformative. |
dual_arm_0008 | The robotic gripper uses a rag to push the trash into the black waste bin. | robotic grippers | rag | A1,B2 | 14 | Both frames show the same two-jaw robotic grippers with identical cylindrical housings and clamp geometry on the robot arms. Only orientation and slight motion blur differ; no structural changes or switch to a hand are present. | The rag is present and matches the reference in type and color, held over the bin. However, the generated frame shows slight motion blur/elongation and a bright streak, indicating minor visual artifacts without major transformation. |
dual_arm_0009 | The right robotic gripper opened the wardrobe and hung the pink hanger with pants inside. | robotic grippers | pink hanger | B1,B2 | 13 | The right frame shows the same two‑jaw rigid robotic gripper with similar housing and linkage as the reference, only rotated and partially cropped. Minor blur/occlusion is present, but structure and type remain consistent. | The pink hanger is clearly present with correct color and general shape. Orientation changes and the hook becomes visible. There’s slight deformation/blur near the right gripper, indicating minor artifacts, but overall it remains consistent with the reference. |
dual_arm_0010 | The robotic gripper uses the key to open the door. | robotic grippers | key | A1,B2 | 14 | Both frames show the same dual-prong robotic grippers with identical structure and tool attachment. Only orientation and position change slightly, with no structural alterations or transformation into a different gripper or hand. Visual clarity is high and consistent across frames. | The key appears in both frames. In the right frame it’s inserted near the lock and slightly blurred/shorter-looking, but still recognizable as the same metallic key. Orientation and position changes account for most differences; only minor visual artifacts are present. |
dual_arm_0011 | The robotic gripper reach into the wardrobe and grab the red shirt. | robotic grippers | red shirt | A1,D2 | 10 | Both frames depict the same rigid two‑jaw robotic gripper with black prongs and a white pad. Structure and design match; only orientation and position vary. No switch to a hand or major shape changes. | The right frame’s red shirt changes from a plain short-sleeve tee to a collared, draped blouse with decorative details, indicating a major transformation beyond orientation or position. |
dual_arm_0012 | The robotic gripper picks up the steak from the plate. | robotic grippers | steak | B1,C2 | 11 | Both frames show the same types of robotic grippers: multi‑prong rigid grippers and a two‑jaw gripper on the white arm. Structures are consistent; only slight blur/angle changes appear, without transformation into a different gripper or a hand. | Both frames show steak pieces—one near the bowl and one on the pink plate. In the right frame, the plate steak is noticeably larger/different in shape, and the held piece looks altered versus the reference. These are clear inconsistencies, though the object remains identifiable as steak. |
dual_arm_0013 | The robotic gripper picks up the green fruit from the beige plate. | robotic grippers | green fruit | D1,A2 | 10 | The reference shows a parallel-jaw style gripper (two rigid jaws), while the generated frame shows a three-pronged claw gripping the object. The gripper type and structure change significantly, not just orientation or minor details. | The green fruit appears in both frames inside the transparent bowl. In the generated frame it’s being grasped but retains the same color and shape. Changes in position/orientation are acceptable; no artifacts, duplication, or major distortions are observed. Visual clarity is good. |
dual_arm_0014 | The robotic gripper picks up the pink object into the dark-colored bowl containing red fruits. | robotic grippers | pink object | B1,A2 | 14 | Both frames show the same robotic grippers: a three-pronged claw on the right arm and a parallel-jaw gripper on the left. Structures match; only orientation and minor occlusion/blur appear in the generated frame. No change to another device type. | The pink object remains present with consistent color and rounded shape. It moved from the transparent bowl to the teal bowl (position change acceptable) and is slightly occluded, but it looks clear and unchanged, with no artifacts or duplication. |
dual_arm_0015 | The robotic gripper picks up the red bottle from the dark-colored container. | robotic grippers | red bottle | A1,A2 | 15 | Both frames show the same multi-pronged robotic grippers with identical structure. Differences are only in orientation and position, with no transformation into another type or noticeable structural change. | The red bottle appears in both frames with the same red body and white cap. It’s simply moved and grasped by the robot, with slight occlusion, but no deformation or artifacts. Overall it is clear and consistent with the reference. |
dual_arm_0016 | The robotic gripper pick up the pink ball from the purple plate and place it in the blue container. | robotic grippers | pink ball | A1,B2 | 14 | Both frames show the same four-jaw rigid robotic grippers with matching frame shape and structure. Differences are only in orientation and position; no structural changes, distortions, or replacement by a hand are evident. | The pink ball appears in both frames, retaining its spherical shape and pink color. Its position changes, which is acceptable. Minor differences in lighting/saturation and apparent size are present but do not significantly affect consistency. |
dual_arm_0017 | The robotic gripper pick up the red bottle from the beige container and place it inside the green bowl. | robotic grippers | red bottle | B1,B2 | 13 | Both frames depict the same rigid multi-pronged robotic grippers. The structure and type match; differences are limited to orientation and partial occlusion with slight artifacts. No switch to a robotic or human hand. | The red bottle is present and looks like the same object (red body, white cap). It has moved to the left bowl and is tilted, partially occluded by the robot/gripper, making it slightly less clear, but overall consistent with the reference. |
dual_arm_0018 | The robotic gripper move the steak from the pink plate to the white plate. | robotic grippers | steak | A1,D2 | 10 | Both frames show the same rigid, multi-pronged robotic grippers. Their structure and shape are consistent; only orientation and position change while grasping the food. No swap to a hand or different end-effector, and no noticeable structural distortions. | In the reference, a single steak sits on the pink plate. In the generated frame, a steak remains on the plate while another steak-like piece is held by the gripper, indicating duplication/artifact rather than mere repositioning or orientation change. |
dual_arm_0019 | The robotic gripper lifted the pink plate containing food. | robotic grippers | pink plate containing food | C1,B2 | 11 | Both frames show robotic grippers (two-jaw type), not hands. However, the generated frame’s grippers differ noticeably in design: larger metallic housings, different jaw shapes and colors compared to the black reference grippers. Orientation changes are fine, but structural differences are clear. | The pink plate appears with similar color and shape, and the food items (rolled piece and meat) match the reference. The generated frame is slightly blurred and the plate seems a bit larger, but these are minor visual issues without major changes. |
dual_arm_0020 | The robotic gripper picks up the red marker one the table and places it inside the pen holder compartment. | robotic grippers | red marker on the table | B1,E2 | 7 | The primary black robotic gripper appears in both frames with the same pronged, rigid structure; differences are mainly orientation and slight rendering. A white arm appears but still uses a gripper, not a hand. Minor visual variations, overall consistent type and geometry. | In the reference frame, a fully red marker lies on the table. In the generated frame, that red marker is absent on the tabletop; only a white marker with a red cap is present. Hence, the red marker is missing. |
dual_arm_0021 | The robotic hand picks up the carton of orange juice on the table and pours it into the white cup. | robotic hands | carton of orange juice | D1,B2 | 9 | The reference shows two multi-finger robotic hands. In the generated frame, the right-side hand becomes a bulky gripper/end-effector rather than a multi-finger hand, while the left hand remains similar. This structural transformation makes the robotic hands highly inconsistent. | The orange juice carton appears in both frames with matching color and design. In the right frame it’s held by the robot and slightly warped/blurred, but still clearly the same carton. Minor visual artifacts only. |
dual_arm_0022 | The robotic hand places the cube onto the upper wooden platform. | robotic hands | cube | D1,B2 | 9 | The right frame replaces the right-side multi‑finger robotic hand with a different clamp/arm mechanism. Its structure and components differ clearly from the reference, beyond orientation changes. Left hand looks similar, but the right hand’s transformation makes the consistency highly inconsistent. | The cube is present in both frames with matching colors (orange top, green/yellow sides) and cubic shape. In the right frame it’s farther and slightly blurred with minor edge/occlusion artifacts from the gripper, but no major deformation. |
dual_arm_0023 | The robotic hand moves the red apple onto the second-level wooden platform. | robotic hands | red apple | A1,B2 | 14 | Both frames show the same setup: an anthropomorphic robotic hand on the left and a two-jaw robotic gripper on the right. Only orientations change as the gripper approaches the apple; the structures and types remain consistent and clear. | The red apple is present and appears similar in color, shape, and size. It moves to the lower compartment and is partially occluded by the robot hand, with minor lighting/perspective differences. No duplication or major artifacts observed. |
dual_arm_0024 | The robotic hand grasps the yellow bottle on the upper shelf and places it onto the table. | robotic hands | yellow bottle | B1,B2 | 13 | Both frames show anthropomorphic multi‑finger robotic hands. The generated frame keeps the same hand type and structure while changing pose to grasp the bottle. Some finger details are slightly less clear due to minor blur/artifacts, but overall consistency remains. | The yellow squeeze bottle appears in both frames with consistent color and shape. In the right frame it’s moved and partially held by the robot, showing slight motion blur/occlusion. Orientation/position changes are acceptable, so only minor visual issues are present. |
dual_arm_0025 | The robotic gripper rinse the plate under the faucet in the sink. | robotic grippers | plate | C1,B2 | 11 | The sink-side robotic gripper changes shape: the left frame shows a compact, rounded black gripper, while the right frame shows a boxy parallel-jaw module with visible wiring. Both are grippers but with noticeable structural differences. The dishwasher-side gripper stays similar. | The plate remains the same circular dish with a blue rim held by the robot in the sink. In the generated frame, running water causes blur/foam that obscures inner details, but the plate’s shape and position are consistent. Minor visual degradation only. |
dual_arm_0026 | The robotic gripper retrieves the red juice carton from the refrigerator. | robotic gripper | red juice carton | B1,A2 | 14 | Both frames show the same two‑jaw black robotic gripper with identical mount and cabling. In the right frame, the red carton partly covers the jaws and the angle shifts, causing minor occlusion/visual differences but no structural change. | The red juice carton appears in both frames with matching color, shape, and size. In the right frame it’s simply being held/rotated by the robot, with no notable artifacts or deformations. Only orientation and position change. |
dual_arm_0027 | The robotic gripper closes the refrigerator door. | robotic grippers | refrigerator door | A1,A2 | 15 | Both frames show the same two-prong robotic grippers on each arm. The structure and number of rigid jaws are consistent, with no switch to a robotic hand. Differences are limited to orientation and reflections, not shape or type. | The refrigerator door in the generated frame matches the reference in material, color, seams, and reflections. It’s simply closed instead of open, which is an acceptable change in orientation/position. No notable artifacts or structural inconsistencies are visible. |
dual_arm_0028 | The robotic gripper open the refrigerator door. | robotic grippers | refrigerator door | A1,A2 | 15 | Both frames show the same two-jaw black robotic grippers with matching shape and structure. Differences are only in orientation and position while opening the fridge and grasping an item. No transformation into a different gripper type or hand is observed. | The refrigerator door in the generated frame is the same unit, simply opened. Material, color, hinge side, and proportions match the reference with no evident artifacts. Orientation change is acceptable; details remain clear and coherent. |
dual_arm_0029 | The robotic gripper insert bread slices into the toaster. | robotic grippers | bread slices | C1,A2 | 12 | Right frame keeps a similar right gripper, but the left-side gripper from the reference is replaced by a different model with distinct body/jaw design and an extra arm. These are structural changes, not just orientation differences. | The bread slices appear as the same brown sliced bread. In the right frame, one slice is being inserted into the toaster while others remain in the basket. Shapes, color, and texture are consistent without artifacts or distortions. |
dual_arm_0030 | The robotic gripper removes the bowl from the sink and places it onto the dish rack on the left. | robotic grippers | bowl | B1,A2 | 14 | Both frames show two-jaw robotic grippers, not robotic hands. Shape and mechanism are consistent; orientation differs as the plate is being placed. Minor visual differences in jaw size/attachments and housing details exist, likely generation artifacts, but overall the grippers match. | The beige bowl appears clearly in both frames with consistent shape and color. Differences are mainly in position/orientation and lighting, which are acceptable. No noticeable artifacts, duplication, or major deformation are present. |
dual_arm_0031 | The robot picks up the black mouse from the open drawer. | robotic grippers | black mouse | D1,B2 | 9 | The left-side robotic gripper changes from a parallel-jaw style to a rounded/suction-like end effector in the generated frame, indicating a different gripper type. Although the right-side gripper remains similar, this structural transformation makes the grippers highly inconsistent overall. | The black mouse appears in both frames. In the generated frame it’s identifiable but slightly occluded by the robot gripper and shows minor blur/shape deviation. Position/orientation changes are fine; overall it stays consistent with the reference with small visual issues. |
dual_arm_0032 | The robotic gripper uses the plastic hammer to hit the pegs on the toy bench. | robotic grippers | plastic hammer | C1,A2 | 12 | Both frames show two-jaw robotic grippers, not hands. However, in the right frame the right gripper’s jaw shape and housing differ noticeably (sleeker casing, different prong geometry/LED), indicating structural changes beyond minor artifacts, though it remains a gripper. | The plastic hammer appears in both frames with the same light handle and white cylindrical head, held by the robot. Only minor orientation/position differences are visible, with no artifacts or duplications. It is clear and consistent. |
dual_arm_0033 | The robotic hand places the bok choy inside the frying pan. | robotic hands | bok choy | A1,B2 | 14 | Both frames depict the same robotic setup: a multi‑finger robotic hand on the left arm and a two‑jaw gripper on the right arm. The structures and features remain consistent; only positions/orientations change as the gripper picks up the vegetable. No transformation or human hand appears. | The bok choy remains the same vegetable with green leaves and white stem, now moved into the pan and held by the robot. It’s partially occluded and slightly blurred, but overall consistent with the reference without major deformation or artifacts. |
dual_arm_0034 | The robotic hand picks up the red wrench near the pink bucket. | robotic hands | red wrench | A1,B2 | 14 | Both frames display the same anthropomorphic robotic hand with five articulated fingers on the left side; finger count, shape, and hardware modules match. Only orientation/pose differs slightly. The gripper on the right is the same type across frames and doesn’t affect hand consistency. | The red wrench appears in both frames with the correct color and general shape. In the generated frame it’s held by the robot, causing partial occlusion and slight perspective/size differences, but no major distortions. Thus it’s mostly consistent with minor visual issues. |
dual_arm_0035 | The robotic hand opens the doors of the black cabinet. | robotic hands | black cabinet | B1,B2 | 13 | Both frames show the same multi‑finger robotic hand; structure and design remain consistent. Differences are limited to pose, perspective, and minor blur/occlusion. No transformation into a gripper or human hand. | The black cabinet remains present with matching shape, color, and ribbed door/handles. The door is opened in the right frame, but there is slight warping and inconsistent panel lines/textures, indicating minor visual artifacts while overall consistency remains. |
dual_arm_0036 | The robotic hand closes the door of the white box. | robotic hands | white box | D1,C2 | 5 | The reference shows an anthropomorphic robotic hand with multiple articulated fingers. In the generated frame, the end-effector becomes a two-pronged parallel gripper on the handle. This is a clear structural change in type, not just orientation or minor artifact. | While the box’s size, color, and top holes match, the generated frame shows a closed front panel with a handle, whereas the reference shows an open cavity. This is a noticeable appearance change beyond orientation/position differences. |
dual_arm_0037 | The robotic hand picks up the orange persimmon from the table and place it inside the white box. | robotic hands | orange persimmon | C1,B2 | 11 | Left frame shows a multi‑finger robotic hand. In the right frame, the end‑effector looks like a blocky gripper with no visible articulated fingers. This indicates a noticeable change in shape/structure beyond mere orientation or position. | The orange persimmon appears in both frames and remains the same object, moved into the box. It’s slightly blurred with a more reddish hue, and the green calyx isn’t visible due to angle/occlusion, but overall it’s consistent with minor visual issues. |
dual_arm_0038 | The robotic hand places the lemon inside the top wooden tray. | robotic hands | lemon | D1,B2 | 9 | The left frame shows multi‑finger robotic hands, while the right frame shows a two‑jaw gripper/clamp. This is a different end‑effector type, indicating a clear structural transformation rather than minor variation. | The lemon appears in both frames. In the right frame it’s being grasped by the robot, retaining the yellow color and general shape. There is slight occlusion/deformation and a size change due to perspective and grip, but only minor visual issues overall. |
dual_arm_0039 | The robotic hand moves the red wrench to the left of the red cup. | robotic hands | red wrench | D1,C2 | 5 | Left frame shows anthropomorphic robotic hands with multiple articulated fingers. Right frame replaces one with a large two‑jaw gripper interacting with the mug. This is a clear transformation from hand to gripper, indicating high inconsistency. | The red wrench appears near the mug but is heavily occluded by the robot/gripper. Its shape is unclear and the blue head isn’t distinctly visible, causing noticeable inconsistencies despite acceptable position/orientation changes. |
dual_arm_0040 | The robotic hand places the tennis ball on the table. | robotic hands | tennis ball | B1,A2 | 14 | Both frames show the same multi‑finger robotic hand at lower-left and a two‑jaw robotic gripper holding the ball. Structures match; differences are limited to angle and minor visual artifacts in the generated frame. | The tennis ball remains clearly visible and consistent in color, texture, and size. It’s simply moved from the plate to the robot gripper, with no artifacts or transformations. |
dual_arm_0041 | The robotic hand places the tennis ball inside the pink bowl. | robotic hands | tennis ball | D1,A2 | 10 | Left frame shows an anthropomorphic robotic hand with multiple fingers. Right frame shows a two-jaw robotic gripper holding the ball. The end-effector type changes from hand to gripper, indicating a major structural transformation, not just orientation or position. | The tennis ball is clearly visible in both frames with consistent color, size, and texture. Only its position changes as the gripper lifts it toward the bowl; no artifacts or distortions. |
dual_arm_0042 | The robotic hand places the orange fruit on the lower wooden tray. | robotic hands | orange fruit | C1,A2 | 12 | The reference shows a slim two-jaw gripper, while the generated frame has a bulkier rectangular end-effector with different jaws/pads. The shape and structure of the robotic hand/gripper change noticeably, beyond minor artifacts, though it remains a gripper. | The orange fruit appears in both frames with consistent color, shape, and size. Only its position changes as it’s being placed into the crate; no artifacts or duplication. |
dual_arm_0043 | The robotic hand picks up the black stapler from the table. | robotic hands | black stapler | B1,E2 | 7 | The multi-finger robotic hand at the lower-left appears in both frames with the same articulated-finger design. Only orientation and minor rendering differences are present. The device near the bowl is a gripper, not the hand, and doesn’t affect the hand’s consistency. | In the reference frame, a black stapler lies to the right of the bowl on the green surface. In the generated frame, that stapler is absent; the spot is occupied by the robot arm instead. Hence the stapler is missing. |
dual_arm_0044 | The robotic hand picks up the blue mug from the lower shelf. | robotic hands | blue mug | C1,C2 | 6 | The right frame introduces a two‑prong gripper with a blue-lit head, while the reference shows an anthropomorphic multi‑finger robotic hand. This indicates a structural change in the robotic hands beyond mere orientation or position differences. | In the right frame, a blue cup-like object is held by the robot, but the handle is not visible and the rim appears unnaturally bright, resembling an artifact. These noticeable visual differences from the reference mug reduce consistency. |
dual_arm_0045 | The robotic hand picks up the Rubik's cube from the table. | robotic hands | Rubik's cube | D1,B2 | 9 | The reference shows an articulated multi-finger robotic hand. In the generated frame, one manipulator becomes a claw-like gripper holding the cube, a different end-effector type, while another hand remains. This is a clear transformation in type, not just orientation. | The Rubik’s cube appears in both frames with correct shape and colored stickers. In the generated frame it’s held by the robot, slightly occluded and a bit blurred, causing minor visual imperfections but overall consistent. |
dual_arm_0046 | The robotic hand picks up the donut from the table and place it into the upper wooden shelf. | robotic hands | donut | D1,B2 | 9 | Left frame shows an anthropomorphic robotic hand with multiple articulated fingers. Right frame shows a different end-effector: a two-jaw robotic gripper holding the object. This is a transformation from hand to gripper, not merely a positional change. | The donut appears in both frames with similar color, shape, and hole. In the right frame it’s being grasped by the robot, causing slight deformation/lighting change, but it remains recognizably the same donut. |
dual_arm_0047 | The robotic hand picks up the white pen with a blue cap and places it into the black mesh pen holder. | robotic hands | white pen with a blue cap | D1,C2 | 5 | Left frame shows an anthropomorphic robotic hand with multiple fingers. Right frame shows a two‑jaw gripper with a different structure. This is a clear transformation to another type of robotic hand. | The right frame shows mainly a blue circular top clamped by the robot, with the white body largely hidden or morphed, making the object appear bulkier than the slim white pen with a blue cap in the left frame. This is a noticeable inconsistency, though not missing. |
dual_arm_0048 | The robotic gripper picks up the red cylindrical can and places it into the plastic bag. | robotic grippers | red cylindrical can | B1,B2 | 13 | Both frames depict the same parallel‑jaw robotic gripper with rigid prongs, not a multi‑finger robotic hand. The structure and type match; only minor visual differences likely from angle/compression and contact with the can are observed. | The red cylindrical can appears in both frames with matching shape, color, and label motifs. The right frame shows it being grasped, with slight differences in detail/sharpness and orientation, but no major deformation or artifact. |
dual_arm_0049 | The robotic gripper holds the menu folder to pass it to the person. | robotic grippers | menu folder | A1,B2 | 14 | Both frames show the same two-prong robotic grippers with identical black jaws and cylindrical bases. Only orientation/viewpoint changes; no structural alterations or transformation to a hand. | The menu folder remains the same maroon cover with gold cursive text and a white tab at the bottom corner. Orientation and lighting differ, with slight blur/occlusion, but overall it matches the reference with only minor visual issues. |
dual_arm_0050 | The robotic gripper receives the notebook handed over by the human. | robotic grippers | notebook | A1,B2 | 14 | Both frames show the same dual parallel-jaw robotic grippers with identical housings and jaws. The right frame only changes position to grasp the object; no structural changes or transformation into a hand are present. | The notebook appears in both frames as the same rectangular item near the person. In the generated frame it’s being lifted by the robot and looks slightly different in color/lighting and thickness, but remains recognizable as the same object. |
dual_arm_0051 | The robotic gripper hands the notebook to the person in front. | robotic grippers | notebook | A1,B2 | 14 | Both frames show the same two-prong robotic grippers with matching structure and materials. Differences are only in orientation and lighting; no change to another gripper type or to a robotic/human hand. | The notebook appears in both frames with the same rectangular shape and dark color. Differences are due to perspective and lighting glare, causing slight visual changes but no major alterations or artifacts. |
dual_arm_0052 | The robot moves forward and picks up the bottle from the table. | robotic grippers | bottle | B1,B2 | 13 | Both frames show two parallel-jaw robotic grippers. The type and structure are consistent, with only minor visual differences in color, detailing, and tip shape. Orientation and position vary but are acceptable. No switch to a robotic hand or human hand. | The red-capped plastic bottle appears in both frames and is recognizable. In the generated frame it shows minor issues—slightly altered size/label clarity and subtle blur—likely from viewpoint or generation artifacts. No duplication or major transformation. |
dual_arm_0053 | The robotic gripper picks up the beverage bottle in the center of the table. | robotic gripper | beverage bottle | D1,B2 | 9 | The right frame shows a different gripper type: the left is a black three‑finger claw, while the right appears as a two‑jaw clamp on a white arm holding the bottle. This is a clear transformation in gripper design, beyond minor visual issues. | The same bottle appears with matching color and shape. It’s being grasped, so orientation changes and part is occluded, with slight blur/lighting differences. These are minor visual issues; overall consistency remains. |
dual_arm_0054 | The robotic gripper organizes the clipboard into the blue container. | robotic grippers | clipboard | B1,B2 | 13 | Both frames show the same dual black robotic grippers with identical multi-pronged jaws. Their orientation changes slightly, but the structure and type remain unchanged. Minor visual differences (angle/lighting and partial occlusion by another arm) exist, so consistency is high but not perfect. | The clipboard appears in both frames as a white board with a metal clip. In the right frame it’s being held and shows slight thickness/clip distortion and reduced detail, but remains recognizably the same object. Differences are minor and due to orientation or generation artifacts. |
dual_arm_0055 | The robotic gripper grasps the orange-yellow toy truck. | robotic grippers | orange-yellow toy truck | C1,B2 | 11 | Both frames show robotic grippers, not hands, but the left gripper’s structure changes noticeably—different base and jaw design—while the right gripper remains similar. This indicates structural inconsistency without replacement by a hand. | The orange-yellow crane truck appears in both frames with matching color and shape. In the right frame it’s being held, causing partial occlusion and slight blur/angle change, but the object remains recognizable and consistent overall. |
dual_arm_0056 | The robotic gripper moves the orange-yellow toy truck closer to the decorative plant in a green vase. | robotic grippers | orange-yellow toy truck | B1,B2 | 13 | Both frames show the same two‑jaw robotic grippers with similar linkage and shape. The generated frame preserves structure but has slight blurring and minor geometry/detail shifts around joints. Orientation changes are expected. Overall mostly consistent with small visual issues. | The orange-yellow toy truck appears in both frames with matching color and general shape. In the right frame it’s repositioned and partly occluded by the robot gripper, causing slight detail loss, but no major deformation or duplication. |
dual_arm_0057 | The robotic gripper presses the button on the cash register. | robotic grippers | cash register | A1,A2 | 15 | Both frames depict the same three-pronged robotic grippers with identical shape and structure. Differences are only in orientation and finger spread due to motion, not in design or type. | The toy cash register appears with the same shape, colors, keypad layout, display, and red scanner attachment. Only a slight viewpoint shift is visible, with no distortions or artifacts. |
dual_arm_0058 | The robotic hand hand over the red coffee cup to the woman. | robotic hands | red coffee cup | A1,A2 | 15 | The robotic hand in the generated frame matches the reference: black glove, five articulated fingers, same white forearm and wrist module with LED; only pose changes. No structural inconsistencies. | The red coffee cup in the generated frame matches the reference: same red body and black lid, similar pattern. Only its position and orientation change as it’s handed over. No duplication, artifacts, or major visual inconsistencies are observed. |
dual_arm_0059 | The robotic hand place the basketball into the mini hoop. | robotic hands | basketball | A1,B2 | 14 | Both frames show the same anthropomorphic robotic hands with black glove-like fingers and identical wrist hardware. Only orientation changes; structure and finger count remain consistent. No switch to grippers or human hands. | The basketball remains red and similarly sized, with only a position change. However, the generated frame shows slight blur/shape distortion and less visible groove lines, indicating minor visual artifacts. Thus it is mostly consistent with small issues. |
dual_arm_0060 | The robotic gripper picks up a yellow bell pepper from the vegetable tray and places it into the shopping cart. | robotic grippers | yellow bell pepper | A1,A2 | 15 | Both frames show the same robotic grippers: a left-side three-jaw claw and a right-side small two-prong pincer on a cylindrical tool. Only orientation/position changes slightly; structure and type remain clear and consistent. | Two yellow bell peppers appear in both frames with matching color, shape, and stems. Their positions and orientation shift slightly, but there are no artifacts or major visual differences. Overall, they are clear and consistent with the reference. |
dual_arm_0061 | The robotic gripper picks up a red apple from the fruit tray and places it into the shopping cart. | robotic grippers | red apple | B1,B2 | 13 | The main end-effector remains a rigid multi-jaw robotic gripper in both frames. Orientation and contact change, and finger visibility/shape shows slight differences, but it’s clearly the same type of gripper without transforming into a hand or another tool. | A clear red apple appears in both frames; in the generated frame it’s being grasped. Color and shape match the reference, though texture and edges look slightly artificial or blurred. Position/orientation changes are acceptable. |
dual_arm_0062 | The robotic gripper picks up a red apple from the wooden crate. | robotic grippers | red apple | A1,B2 | 14 | Both frames show the same black two‑jaw robotic gripper with matching plates, bolts, and prong geometry. Only orientation/pose changes; no structural alterations or hand substitution. Visual clarity remains good, so it is consistent. | The red apple appears in both frames with consistent color and general shape. Orientation and position change due to the gripper, and the texture/shine looks slightly different and smoother in the generated frame, but these are minor visual issues rather than major inconsistencies. |
dual_arm_0063 | The robotic gripper picks up a bunch of grapes from the wooden crate. | robotic gripper | bunch of grapes | A1,B2 | 14 | Both frames show the same two-jaw rigid robotic gripper with identical serrated tips and mounting holes. Differences are only in orientation and lighting; no structural changes. | The grape cluster remains present with similar purple/red berries and stem structure, in the same area near the peaches. Minor differences in lighting and berry count/arrangement are visible, but overall it stays coherent and close to the reference. |
dual_arm_0064 | The robotic grippers place a green packaged item into the shopping cart. | robotic grippers | green packaged item | A1,D2 | 10 | Both frames show the same black two-jaw robotic gripper with identical body and linkage. The generated frame keeps the same structure; only orientation and slight occlusion from holding the green bag differ. No structural changes or switch to a hand. | Appearance matches, but the generated frame shows an extra green packaged item being held while two remain on the shelf. This duplication compared to the reference constitutes a major inconsistency. |
dual_arm_0065 | The robotic gripper places the peanut bag into the shopping cart. | robotic grippers | peanut bag | B1,B2 | 13 | The generated frame shows the same black parallel-jaw robotic grippers as the reference. Structure and type remain consistent; only minor differences in angle, occlusion, and small visual artifacts are present. No switch to a robotic hand or different gripper. | The white peanut bag (bibizan) seen on the shelf in the reference appears in the generated frame inside the cart. Despite being crumpled and partially occluded, its colors and peanut imagery match, showing mostly consistent appearance with minor visual issues. |
dual_arm_0066 | The robotic gripper picks up and places the white bottle with a blue pump dispenser inside the plastic bag. | robotic grippers | white bottle with a blue pump dispenser | B1,B2 | 13 | Both frames show a two-jaw robotic gripper with rigid prongs. The generated frame maintains the same gripper type and structure, though orientation, occlusion by the bag, and added blue pads introduce minor visual differences. | The white bottle with a blue pump is present in the right frame. The blue pump and part of the bottle are visible near the gripper, but it’s partially occluded by the bag and less clear. Colors and form match the reference with minor visibility issues. |
dual_arm_0067 | The robotic gripper selects a yellow tube of chips from the shelf and places it into the shopping cart. | robotic grippers | yellow tube of chips | A1,B2 | 14 | Both frames show the same rigid multi-pronged robotic grippers with identical structure and hardware. Differences are limited to orientation and one gripper closing to hold an item; no structural changes or transformation into a different gripper or hand. | The yellow tube of chips appears in both frames with similar color, shape, and placement. The generated frame shows minor text/label distortions and slight blur, but overall it remains recognizable and consistent. |
dual_arm_0068 | The robotic grippers pick up the purple eggplants and place them inside the shopping cart. | robotic grippers | purple eggplants | A1,B2 | 14 | Both frames show the same skeletal clamp-style robotic grippers with rigid prongs and similar structure. The generated frame keeps the same type and design, only differing in orientation and grasping an item. No transformation to a different gripper or hand. | Purple eggplants appear in both frames, with one being grasped in the right frame. Their overall shape and color match the reference, though the calyx looks more yellow and size differs slightly due to angle. Minor visual deviations but generally consistent. |
dual_arm_0069 | The robotic gripper places the yellow tube of chips inside the plastic bag. | robotic grippers | yellow tube of chips | B1,B2 | 13 | Both frames show the same three-prong robotic gripper with a metal block base. In the right frame, orientation changes and the gripper is partially occluded by the bag, but its structure remains clear and consistent—not a robotic hand. | The yellow chips tube appears in both frames with matching color, shape, and lid design. In the right frame it’s rotated and partially inside the bag, with slight blur/occlusion, but remains largely consistent. |
dual_arm_0070 | The robotic grippers pick a box of Franxxi biscuits from the shelf and place it in the shopping cart. | robotic grippers | Franxxi biscuits | A1,D2 | 10 | Both frames show the same clamp-style robotic grippers with rigid jaws on black arms. The right frame simply shows them closing on the item with slight orientation changes and minor motion blur. No structural changes or switch to a robotic/human hand are observed. | Franxxi biscuits are present, but the right frame shows the pack heavily warped and deformed with smeared text—an obvious AI artifact—unlike the clear rectangular box in the reference. This is a major transformation, not just a minor issue. |
dual_arm_0071 | The robotic gripper places a banana into the shopping cart. | robotic grippers | banana | A1,B2 | 14 | Both frames show the same two-jaw rigid robotic grippers on the arms. The structure and shape match, with only orientation and grasp state changing. No transformation to a different gripper type or to a robotic/human hand. | The bananas in the tray look the same in shape and color as the reference. One banana is being held by the robot and is slightly blurred/occluded, but overall appearance remains consistent without major deformation or artifacts. |
dual_arm_0072 | The robotic gripper picks up the Danisa cranberry cookies box and places it in the shopping cart. | robotic grippers | Danisa cranberry cookies box | B1,C2 | 11 | Both frames show rigid two‑jaw robotic grippers with similar black housings. The generated frame keeps the same type and structure, though the right gripper has slight geometric/visual artifacts and minor blur, indicating small inconsistencies but no transformation into a different end‑effector. | The Danisa cranberry cookies box appears in both frames, but the generated frame shows noticeable distortions in text and graphics, with altered details and blur. It’s identifiable yet clearly inconsistent with the reference. |
dual_arm_0073 | The robotic gripper picks up a blue box of snack and places it into the shopping cart. | robotic grippers | a blue box of snack | A1,C2 | 12 | Both frames show the same dual robotic grippers with rigid multi-prong jaws and metallic rectangular tips. Structure, components, and design match; only orientation and the act of holding an item differ. No switch to a robotic hand or structural change is evident. | The blue snack box is present in both frames, but the generated frame shows different packaging details and shape/size (appears more like a different blue pack). These noticeable visual differences reduce consistency, though the item isn’t missing or duplicated. |
dual_arm_0074 | The robotic gripper picks up the green Franxxi biscuit box from the shelf. | robotic grippers | green Franxxi biscuit box | C1,E2 | 3 | Both frames show robotic grippers, but the generated frame exhibits noticeable structural changes: the lower gripper’s jaws are distorted and reshaped, and the upper gripper appears to shift from a multi-pronged claw to a two-jaw clamp. Orientation changes aside, shapes differ clearly. | The bright green Franxxi biscuit box visible in the reference frame is not present in the generated frame; it’s replaced by darker Franzzi boxes and other items. Thus the specific green box is missing. |
dual_arm_0075 | The robotic gripper places the Lay's black bags into the blue shopping basket. | robotic grippers | Lay's black bags | B1,C2 | 11 | Both frames show similar robotic grippers with rigid jaws and round housings. Orientation changes, but design is preserved. The right frame has slight blur/minor jaw detail differences, yet no structural change or replacement by a hand. | In the reference, two Lay’s black bags are clearly visible in the blue bin. In the generated frame, only one Lay’s black bag appears. The design looks similar, but the quantity change is a noticeable inconsistency. |
dual_arm_0076 | The robotic arm folds the orange paper. | robotic arms | orange paper | A1,B2 | 14 | Both frames show the same dual ABB robotic arms with identical white/gray multi-jointed structure and two-finger grippers. Only minor pose changes while holding the orange paper are visible; no structural alterations or substitutions occur. | The orange paper appears in both frames and is held by the robot. Color and identity match, but the right frame shows slight motion blur/shape deformation, making it less clear. Orientation changes are acceptable. Overall, mostly consistent with minor visual issues. |
dual_arm_0077 | The robotic arms assemble parts on the production line. | robotic arms | assembly parts | A1,B2 | 14 | The generated frame preserves the same dual-arm ABB robots with identical shape, joints, and two-pronged grippers. Only arm poses and orientations differ due to motion; no structural changes or hand substitution are visible. | Assembly parts remain present and similarly shaped and arranged. Minor motion blur and slight lighting/color differences are visible, but no missing components, duplication, or major artifacts. |
dual_arm_0078 | The robotic arms pull the levers on both sides of the production line. | robotic arms | levers | A1,B2 | 14 | Both frames show the same dual ABB-style robotic arms with grey-white segmented joints and small parallel grippers. Only the poses differ; structure, number of joints, and grippers remain consistent and clear, with no transformation or replacement observed. | The levers are present in the right frame, matching the reference in count and general placement. There are minor shape/thickness differences and slight rendering artifacts, but no major transformation or duplication. Orientation and position shifts are acceptable. |
dual_arm_0079 | The robotic arms picks up a test tube from the rack. | robotic arms | test tube | A1,A2 | 15 | Both frames show the same dual segmented robotic arms with parallel-jaw grippers. The structure, color, and end-effector type remain consistent; only pose and orientation change as the right arm grips a tube. No transformation into a different arm or hand is observed. | Test tubes retain the same yellow caps, transparent bodies, and rack arrangement. One is grasped by the robot with only orientation/position changes. Visual details remain clear without artifacts or duplication; minor lighting differences only. |
dual_arm_0080 | The robotic arms adjusts its positioning mechanism with the orange fixtures. | robotic arms | orange fixture | A1,B2 | 14 | Both frames show the same dual segmented robotic arms with black grippers holding orange objects. Structure and gripper type remain unchanged; only pose/orientation varies, which is acceptable. The arms are clear and consistent. | The orange fixtures appear on both arms with the same color and general shape. Orientation changes are expected. Minor visual differences—slight contour deformation and smudged markings—are visible, but overall they remain consistent with the reference. |
dual_arm_0081 | The robotic arms folds the piece of fabric on the table. | robotic arms | fabric | A1,A2 | 15 | The robotic arm in the right frame matches the left: same black multi-jointed structure, cyan light strip, and two-prong gripper. Only orientation and position differ while manipulating the cloth; no structural or type change observed. | The navy fabric appears the same material and color; only its folding state changes due to manipulation. No artifacts, duplication, or loss. Consistency is clear. |
dual_arm_0082 | The robotic arms folds the blue shirt on the table. | robotic arms | blue shirt | B1,B2 | 13 | Both frames show the same pair of black robotic arms with cyan light strips and similar gripper-style end effectors. Orientation changes, but structure remains consistent. Minor visual softness/artifacts around joints and grippers appear in the generated frame, without structural changes or hand substitution. | The blue shirt is present in both frames with consistent color and identity. In the right frame it’s slightly lifted and wrinkled by the robot, with minor motion/shape distortions, but it remains clearly the same garment. |
dual_arm_0083 | The robotic arms places the black object and the red object back into their respective boxes.. | robotic arms | red cap | C1,B2 | 11 | The cobot arm bodies remain similar, but the end effectors change noticeably: the left frame shows a blue tri‑pronged gripper, while the right frame shows different white/black grippers. This structural change in the gripper is significant, though it’s still a gripper, not a hand. | Red caps appear in both frames with similar count, color, and placement. The generated frame shows slight softness and minor edge/deformation artifacts, but overall the caps remain recognizable and consistent. |
dual_arm_0084 | The robotic arms places the rectangular and cylindrical components into their designated slots on the trays. | robotic arms | black component | A1,B2 | 14 | The robotic arms and their two-jaw grippers remain the same in design and structure. Only positions/orientations change between frames. No switch to a robotic hand or human hand, and no structural deformation is evident. | The black rectangular components appear in both frames with similar quantity and placement. Minor visual differences (slight blurring and small shifts) are present but do not alter the component’s identity, making it mostly consistent with the reference. |
dual_arm_0085 | The robotic arms pours liquid from the blue container into the test tubes. | robotic arms | blue container | A1,A2 | 15 | Both frames show the same dual articulated grey-white robotic arms with consistent shape and end-effectors. Differences are limited to slight pose/orientation; no structural change or substitution with a hand. The arms are clear and match the reference. | The blue container appears in both frames with the same beaker shape, blue liquid, and position within the clear holder; only minor shifts or lighting differences. |
dual_arm_0086 | The robotic arms uses a pipette to transfer liquid into the test tubes. | robotic arms | pipette | B1,B2 | 13 | Both frames show the same dual robotic arms with identical grey-white links, joints, and tool holding a pipette. Structure and type are consistent; only minor differences in pose and small detail/blur are visible in the generated frame. | The pipette is present and identifiable with similar form and placement. Branding and color band appear, though slightly blurred, and the tip looks shorter/thicker than in the reference. These are minor visual artifacts; overall consistency is maintained. |
dual_arm_0087 | The robotic arms picks up the red token labeled '4' from the table. | robotic arms | red token labeled '4' | B1,B2 | 13 | Both frames show the same black articulated robotic arms with blue light strips and identical bases. The generated frame changes only the arms’ poses, with slight rendering/blur artifacts at the end-effectors. Overall structure remains consistent, with minor visual issues. | The red token labeled '4' is present in the right frame with matching color and shape. The number remains visible, though slightly blurred/occluded and shifted due to motion/perspective. These are minor visual issues; overall consistency is maintained. |
dual_arm_0088 | The robotic arm picks up the pink bowl from the checkerboard platform. | robotic arms | pink bowl | A1,B2 | 14 | Both frames show the same black robotic arms with blue light strips and clamp-style grippers. Only orientation/position changes; structure and gripper type remain consistent. Any differences are minimal and due to perspective or motion. | The pink bowl is present with matching color and general shape, now held by the robot. Orientation/position change is acceptable. There are slight size/shape distortions and minor blur/artifacts around the rim/gripper, but overall it remains consistent. |
dual_arm_0089 | The robotic arms places the cube from the table into the yellow bowl. | robotic arms | cube | A1,B2 | 14 | Both frames show the same black articulated robotic arms with blue illuminated panels and the same gripper type; the end effector still grips the cube. Only slight pose shifts; no structural changes or replacement by a hand. | The multicolored cube is present and recognizable, with similar color pattern and shape. It’s moved/held by the robot, but shows minor visual artifacts (slight blur/deformation of squares). These are small issues, not major inconsistencies. |
dual_arm_0090 | The robotic arms lifts up the yellow corn and the apple from the table. | robotic arms | yellow corn&apple | B1,B2 | 13 | Both frames show two black robotic arms with blue-lit segments and similar multi-prong robotic grippers. The right frame mainly changes position and action (grasping an apple), with slight artifact/shape noise at the gripper. No switch to a different arm type or to a hand. | The yellow corn and apples appear in both frames. The corn is the same object with slight rotation. Two apples remain, one being grasped. Minor visual artifacts/shape changes on the grasped apple are present, but overall consistency is maintained. |
dual_arm_0091 | The robotic arms turn on the faucet and fill the glass with water. | robotic arms | glass | A1,C2 | 12 | Both frames show the same two articulated robotic arms with identical structures and gripper types; only their orientations shift slightly. No change in end-effector type or transformation into a different hand/arm is observed. Minor lighting/motion differences don’t affect consistency. | The glass remains present and similar, but in the generated frame it shifts to the sink edge and shows an odd black handle/holder artifact, with the original holder appearing empty. These geometry/attachment inconsistencies are noticeable though not a full transformation. |
dual_arm_0092 | The robotic arms rinses the plate under running water. | robotic arms | plate | A1,C2 | 12 | Both frames show the same dual grey robotic arms with white coverings and orange two-pronged grippers. The structure and design match; only orientation and pose change slightly. No transformation or replacement is observed. | The plate is present and similarly shaped, but the right frame lacks the orange ring design seen on the reference plate, appearing as a plain white plate. This is a noticeable visual inconsistency, though the object remains a plate. |
dual_arm_0093 | The robotic arm prepares to fold the green sweater. | robotic arms | green sweater | A1,A2 | 15 | Both frames show the same dual robotic arms with identical two-jaw grippers (orange pads, white plates). Structure remains the same; differences are only in orientation/position while manipulating the cloth. No transformation or substitution occurs. | The mint-green knit sweater matches in color, texture, and design. Differences are due to folding by the robot, not artifacts. No duplication or major visual inconsistencies are present. |
dual_arm_0094 | The robotic arms picks up the plate from the drying rack. | robotic arms | plate | B1,B2 | 13 | The two UR-style robotic arms with prong-type grippers remain present and of the same type in both frames. The generated frame shows slight shape/artifact changes around the grippers and coverings, but overall structure and type are consistent. | The plate is present and matches in shape and color with a blue rim, though orientation changes. Minor visual issues include a less detailed rim pattern and slight blurring/distortion compared to the reference. |
dual_arm_0095 | The robotic arm picks up several plastic pellets from the tabletop. | robotic arms | packing peanut | B1,B2 | 13 | Both frames depict the same dual robotic arms with cylindrical segments, LED rings, and a prong gripper. The generated frame shows small orientation shifts and slight motion blur on end-effectors, but no structural change. Thus consistency is high with minor visual issues. | Packing peanuts appear in both frames on the table and inside the bag. Their white color and general shape match. The generated frame shows slight blur and minor differences in clumping/quantity, but overall remains consistent with the reference. |
dual_arm_0096 | The robotic arm grabs the pink toy figure on the blue surface. | robotic arms | pink toy figure | A1,A2 | 15 | Both frames show the same metallic robotic arm with a black multi‑finger robotic hand/gripper. The structure, joints, and end-effector design are unchanged; only the pose varies as it grasps the toy. No transformation or noticeable distortions are present. | The pink starfish toy maintains the same shape, colors (pink body, green shorts), and facial details. Only its orientation and position change as it’s grasped by the robot. There are no distortions, artifacts, or missing features. Clarity remains high. |
dual_arm_0097 | The robotic arms pour water from the bottle into the cup. | robotic arms | bottle | C1,B2 | 11 | The right frame’s robotic arms differ noticeably: longer, bulkier segmented links with different wrist/end-effector housings and visible cabling, unlike the compact identical arms in the reference. They remain gripper-type, not hands, but the structure changes are evident. | The right frame shows the same plastic bottle with white cap and red label, now tilted and held by the robot while pouring. Minor blur/warping and partial occlusion by the gripper reduce clarity, but it remains consistent with the reference overall. |
dual_arm_0098 | The robotic arms fold the pants neatly on the table. | robotic arms | pants | B1,B2 | 13 | Both frames depict the same dual black robotic arms with two-jaw grippers and similar joints. The right frame shows a grasping pose. Minor generative artifacts and slight detail smoothing are present, but overall structure and gripper type remain consistent. | The pants remain the same black shorts with a yellow/orange trim, but they are crumpled and being grasped by the robot. Minor deformation and slight color/lighting artifacts appear, yet the item is clearly consistent overall. |
dual_arm_0099 | The robotic arms fold the clothes neatly on the table. | robotic arms | clothes | B1,B2 | 13 | Both frames depict the same two white robotic arms with similar multi-prong grippers. Structure and type are consistent; only minor differences in gripper openness and positioning while grasping the cloth are visible. | The clothes match in type, color, and stripe pattern. In the generated frame one piece is being lifted, creating wrinkles and a stretched look. This is a minor visual change due to manipulation, not an artifact or major transformation. |
dual_arm_0100 | The robotic hands lift the bottle with both hands. | robotic hands | bottle | C1,B2 | 11 | Both frames show robotic hands, but the generated frame alters finger shapes/count, with more clamp-like, fused digits and structural distortions. These are beyond mere pose changes, indicating noticeable inconsistencies. | The bottle remains present with similar shape, color, and white cap, positioned near the robot’s hands. Minor artifacts/blur and slight label simplification occur in the generated frame, but overall it matches the reference with small visual deviations. |
End of preview.
WEAVE-0.5 — RBench Evaluation Submission
This is the single-checkpoint submission bundle for WEAVE-0.5. It contains the complete set of 650 generated RBench videos and the corresponding GPT-5 evaluation artifacts.
Contents
videos/task/— 250 videos across five task dimensions, 50 per dimension.videos/embodiment/— 400 videos across four robot embodiments, 100 per embodiment.evaluation/gpt/detailed_results_650.csv— one unified row per generated video.evaluation/gpt/score_summary.csv— the nine dimension scores and equal-weight overall score.evaluation/gpt/evaluation_config.md— judge, endpoint, evaluator, and reproducibility information.email_draft.md— the submission email draft.
Published links
- ReVidgen evaluation framework
- Unified 650-row detailed results
- GPT-5 score summary
- GPT-5 evaluation configuration
- Generated videos
GPT-5 results
| Dimension | Score |
|---|---|
| Common Manipulation | 0.5750 |
| Spatial Relationship | 0.6890 |
| Multi-entity Collaboration | 0.4830 |
| Long-horizon Planning | 0.6008 |
| Visual Reasoning | 0.5480 |
| Single Arm | 0.7548 |
| Dual Arm | 0.6896 |
| Quadruped Robot | 0.7130 |
| Humanoid Robot | 0.7272 |
| Overall (9-dimension equal weight) | 0.6422 |
Evaluation
- Generated videos: 650 total (250 task + 400 embodiment).
- Judge model:
gpt-5-2025-08-07. - Judge provider: ChatAnywhere.
- Judge base URL:
https://api.chatanywhere.tech/v1. - Overall score: 0.642247777778 (rounded to 0.6422).
- Aggregation: equal-weight average over the nine RBench dimensions.
The primary submission result is GPT-5. Any secondary judge artifacts, if present, are reference-only and are not part of the reported GPT-5 score.
- Downloads last month
- 210