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
doc_id: int64
target: string
filtered_resps: string
token_counts: list<item: struct<input_tokens: int64, output_tokens: int64, reasoning_tokens: int64>>
child 0, item: struct<input_tokens: int64, output_tokens: int64, reasoning_tokens: int64>
child 0, input_tokens: int64
child 1, output_tokens: int64
child 2, reasoning_tokens: int64
doc_hash: string
aqua_strategy_accuracy: struct<id: string, level: int64, pred_level: int64, judged: bool, judge_raw: string, score: double>
child 0, id: string
child 1, level: int64
child 2, pred_level: int64
child 3, judged: bool
child 4, judge_raw: string
child 5, score: double
aqua_judge_rate: struct<id: string, level: int64, pred_level: int64, judged: bool, judge_raw: string, score: double>
child 0, id: string
child 1, level: int64
child 2, pred_level: int64
child 3, judged: bool
child 4, judge_raw: string
child 5, score: double
model_name: string
group_subtasks: struct<aqua_level0: list<item: null>, aqua_level1: list<item: null>, aqua_level2: list<item: null>, (... 30 chars omitted)
child 0, aqua_level0: list<item: null>
child 0, item: null
child 1, aqua_level1: list<item: null>
child 0, item: null
child 2, aqua_level2: list<item: null>
child 0, item: null
child 3, aqua_level3: list<item: null>
child 0, item: null
start_time: double
config: struct<model: string, model_args: string, model_backend: string, batch_size: string, batch_sizes: li (... 1270 chars omitted)
...
e, total_input_tokens: double, total_output_tokens (... 131 chars omitted)
child 0, docs: double
child 1, docs_with_token_counts: double
child 2, total_input_tokens: double
child 3, total_output_tokens: double
child 4, total_tokens: double
child 5, total_correct_score: double
child 6, avg_output_tokens_per_sample: double
child 7, tokens_per_correct_answer: null
child 3, aqua_level0: struct<docs: double, docs_with_token_counts: double, total_input_tokens: double, total_output_tokens (... 131 chars omitted)
child 0, docs: double
child 1, docs_with_token_counts: double
child 2, total_input_tokens: double
child 3, total_output_tokens: double
child 4, total_tokens: double
child 5, total_correct_score: double
child 6, avg_output_tokens_per_sample: double
child 7, tokens_per_correct_answer: null
child 1, overall: struct<docs: double, docs_with_token_counts: double, total_input_tokens: double, total_output_tokens (... 131 chars omitted)
child 0, docs: double
child 1, docs_with_token_counts: double
child 2, total_input_tokens: double
child 3, total_output_tokens: double
child 4, total_tokens: double
child 5, total_correct_score: double
child 6, avg_output_tokens_per_sample: double
child 7, tokens_per_correct_answer: null
fewshot_as_multiturn: bool
lmms_eval_version: string
to
{'results': {'aqua_level0': {'alias': Value('string'), 'aqua_strategy_accuracy,none': Value('float64'), 'aqua_strategy_accuracy_stderr,none': Value('string'), 'aqua_strategy_accuracy_stderr_clt,none': Value('float64'), 'aqua_strategy_accuracy_stderr_clustered,none': Value('string'), 'aqua_judge_rate,none': Value('float64'), 'aqua_judge_rate_stderr,none': Value('string'), 'aqua_judge_rate_stderr_clt,none': Value('float64'), 'aqua_judge_rate_stderr_clustered,none': Value('string')}, 'aqua_level1': {'alias': Value('string'), 'aqua_strategy_accuracy,none': Value('float64'), 'aqua_strategy_accuracy_stderr,none': Value('string'), 'aqua_strategy_accuracy_stderr_clt,none': Value('float64'), 'aqua_strategy_accuracy_stderr_clustered,none': Value('string'), 'aqua_judge_rate,none': Value('float64'), 'aqua_judge_rate_stderr,none': Value('string'), 'aqua_judge_rate_stderr_clt,none': Value('float64'), 'aqua_judge_rate_stderr_clustered,none': Value('string')}, 'aqua_level2': {'alias': Value('string'), 'aqua_strategy_accuracy,none': Value('float64'), 'aqua_strategy_accuracy_stderr,none': Value('string'), 'aqua_strategy_accuracy_stderr_clt,none': Value('float64'), 'aqua_strategy_accuracy_stderr_clustered,none': Value('string'), 'aqua_judge_rate,none': Value('float64'), 'aqua_judge_rate_stderr,none': Value('string'), 'aqua_judge_rate_stderr_clt,none': Value('float64'), 'aqua_judge_rate_stderr_clustered,none': Value('string')}, 'aqua_level3': {'alias': Value('string'), 'aqua_strategy_accuracy,no
...
ken_counts': Value('float64'), 'total_input_tokens': Value('float64'), 'total_output_tokens': Value('float64'), 'total_tokens': Value('float64'), 'total_correct_score': Value('float64'), 'avg_output_tokens_per_sample': Value('float64'), 'tokens_per_correct_answer': Value('null')}, 'aqua_level0': {'docs': Value('float64'), 'docs_with_token_counts': Value('float64'), 'total_input_tokens': Value('float64'), 'total_output_tokens': Value('float64'), 'total_tokens': Value('float64'), 'total_correct_score': Value('float64'), 'avg_output_tokens_per_sample': Value('float64'), 'tokens_per_correct_answer': Value('null')}}, 'overall': {'docs': Value('float64'), 'docs_with_token_counts': Value('float64'), 'total_input_tokens': Value('float64'), 'total_output_tokens': Value('float64'), 'total_tokens': Value('float64'), 'total_correct_score': Value('float64'), 'avg_output_tokens_per_sample': Value('float64'), 'tokens_per_correct_answer': Value('null')}}, 'task_hashes': {'aqua_level3': Value('string'), 'aqua_level2': Value('string'), 'aqua_level1': Value('string'), 'aqua_level0': Value('string')}, 'model_source': Value('string'), 'model_name': Value('string'), 'model_name_sanitized': Value('string'), 'system_instruction': Value('null'), 'system_instruction_sha': Value('null'), 'fewshot_as_multiturn': Value('bool'), 'chat_template': Value('null'), 'chat_template_sha': Value('null'), 'start_time': Value('float64'), 'end_time': Value('float64'), 'total_evaluation_time_seconds': Value('string')}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
return get_rows(
dataset=dataset,
...<4 lines>...
column_names=column_names,
)
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
File "/src/services/worker/src/worker/utils.py", line 483, in safe_iter
yield from ds.decode(False) if ds.features else ds
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2951, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2461, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2486, 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 547, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 430, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
doc_id: int64
target: string
filtered_resps: string
token_counts: list<item: struct<input_tokens: int64, output_tokens: int64, reasoning_tokens: int64>>
child 0, item: struct<input_tokens: int64, output_tokens: int64, reasoning_tokens: int64>
child 0, input_tokens: int64
child 1, output_tokens: int64
child 2, reasoning_tokens: int64
doc_hash: string
aqua_strategy_accuracy: struct<id: string, level: int64, pred_level: int64, judged: bool, judge_raw: string, score: double>
child 0, id: string
child 1, level: int64
child 2, pred_level: int64
child 3, judged: bool
child 4, judge_raw: string
child 5, score: double
aqua_judge_rate: struct<id: string, level: int64, pred_level: int64, judged: bool, judge_raw: string, score: double>
child 0, id: string
child 1, level: int64
child 2, pred_level: int64
child 3, judged: bool
child 4, judge_raw: string
child 5, score: double
model_name: string
group_subtasks: struct<aqua_level0: list<item: null>, aqua_level1: list<item: null>, aqua_level2: list<item: null>, (... 30 chars omitted)
child 0, aqua_level0: list<item: null>
child 0, item: null
child 1, aqua_level1: list<item: null>
child 0, item: null
child 2, aqua_level2: list<item: null>
child 0, item: null
child 3, aqua_level3: list<item: null>
child 0, item: null
start_time: double
config: struct<model: string, model_args: string, model_backend: string, batch_size: string, batch_sizes: li (... 1270 chars omitted)
...
e, total_input_tokens: double, total_output_tokens (... 131 chars omitted)
child 0, docs: double
child 1, docs_with_token_counts: double
child 2, total_input_tokens: double
child 3, total_output_tokens: double
child 4, total_tokens: double
child 5, total_correct_score: double
child 6, avg_output_tokens_per_sample: double
child 7, tokens_per_correct_answer: null
child 3, aqua_level0: struct<docs: double, docs_with_token_counts: double, total_input_tokens: double, total_output_tokens (... 131 chars omitted)
child 0, docs: double
child 1, docs_with_token_counts: double
child 2, total_input_tokens: double
child 3, total_output_tokens: double
child 4, total_tokens: double
child 5, total_correct_score: double
child 6, avg_output_tokens_per_sample: double
child 7, tokens_per_correct_answer: null
child 1, overall: struct<docs: double, docs_with_token_counts: double, total_input_tokens: double, total_output_tokens (... 131 chars omitted)
child 0, docs: double
child 1, docs_with_token_counts: double
child 2, total_input_tokens: double
child 3, total_output_tokens: double
child 4, total_tokens: double
child 5, total_correct_score: double
child 6, avg_output_tokens_per_sample: double
child 7, tokens_per_correct_answer: null
fewshot_as_multiturn: bool
lmms_eval_version: string
to
{'results': {'aqua_level0': {'alias': Value('string'), 'aqua_strategy_accuracy,none': Value('float64'), 'aqua_strategy_accuracy_stderr,none': Value('string'), 'aqua_strategy_accuracy_stderr_clt,none': Value('float64'), 'aqua_strategy_accuracy_stderr_clustered,none': Value('string'), 'aqua_judge_rate,none': Value('float64'), 'aqua_judge_rate_stderr,none': Value('string'), 'aqua_judge_rate_stderr_clt,none': Value('float64'), 'aqua_judge_rate_stderr_clustered,none': Value('string')}, 'aqua_level1': {'alias': Value('string'), 'aqua_strategy_accuracy,none': Value('float64'), 'aqua_strategy_accuracy_stderr,none': Value('string'), 'aqua_strategy_accuracy_stderr_clt,none': Value('float64'), 'aqua_strategy_accuracy_stderr_clustered,none': Value('string'), 'aqua_judge_rate,none': Value('float64'), 'aqua_judge_rate_stderr,none': Value('string'), 'aqua_judge_rate_stderr_clt,none': Value('float64'), 'aqua_judge_rate_stderr_clustered,none': Value('string')}, 'aqua_level2': {'alias': Value('string'), 'aqua_strategy_accuracy,none': Value('float64'), 'aqua_strategy_accuracy_stderr,none': Value('string'), 'aqua_strategy_accuracy_stderr_clt,none': Value('float64'), 'aqua_strategy_accuracy_stderr_clustered,none': Value('string'), 'aqua_judge_rate,none': Value('float64'), 'aqua_judge_rate_stderr,none': Value('string'), 'aqua_judge_rate_stderr_clt,none': Value('float64'), 'aqua_judge_rate_stderr_clustered,none': Value('string')}, 'aqua_level3': {'alias': Value('string'), 'aqua_strategy_accuracy,no
...
ken_counts': Value('float64'), 'total_input_tokens': Value('float64'), 'total_output_tokens': Value('float64'), 'total_tokens': Value('float64'), 'total_correct_score': Value('float64'), 'avg_output_tokens_per_sample': Value('float64'), 'tokens_per_correct_answer': Value('null')}, 'aqua_level0': {'docs': Value('float64'), 'docs_with_token_counts': Value('float64'), 'total_input_tokens': Value('float64'), 'total_output_tokens': Value('float64'), 'total_tokens': Value('float64'), 'total_correct_score': Value('float64'), 'avg_output_tokens_per_sample': Value('float64'), 'tokens_per_correct_answer': Value('null')}}, 'overall': {'docs': Value('float64'), 'docs_with_token_counts': Value('float64'), 'total_input_tokens': Value('float64'), 'total_output_tokens': Value('float64'), 'total_tokens': Value('float64'), 'total_correct_score': Value('float64'), 'avg_output_tokens_per_sample': Value('float64'), 'tokens_per_correct_answer': Value('null')}}, 'task_hashes': {'aqua_level3': Value('string'), 'aqua_level2': Value('string'), 'aqua_level1': Value('string'), 'aqua_level0': Value('string')}, 'model_source': Value('string'), 'model_name': Value('string'), 'model_name_sanitized': Value('string'), 'system_instruction': Value('null'), 'system_instruction_sha': Value('null'), 'fewshot_as_multiturn': Value('bool'), 'chat_template': Value('null'), 'chat_template_sha': Value('null'), 'start_time': Value('float64'), 'end_time': Value('float64'), 'total_evaluation_time_seconds': Value('string')}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
lmms-eval-embodied — evaluation results / 评测结果
Per-question outputs and scores of 10 models on the embodied-intelligence suite in
lmms-eval-embodied (47-item capability taxonomy). Generated with run_model.sh.
10 个模型在 lmms-eval-embodied 具身智能评测套件上的逐题结果与分数。
| Path | Content |
|---|---|
reports/summary_capability_agg.csv |
one row per capability / 每项能力一行 |
reports/summary_capability.csv |
capability × source / 能力 × 来源明细 |
reports/summary_subtasks.csv |
every subtask / 全部子任务 |
reports/README_capability.md |
scoring caliber and known limitations (Chinese) / 计分口径与已知局限 |
results/<model>/<benchmark>/*_samples_<task>.jsonl |
per-question records: doc_id, target, model response, score fields |
results/<model>/<benchmark>/*_results.json |
aggregated metrics and run config |
Notes / 说明
- Question text (
input) is removed; benchmark data is not redistributed. Join ondoc_idwith the data rebuilt byprepare_data.pyin the code repository. 已去掉题干，按doc_id与代码仓库prepare_data.py重建的数据对应。 - When a task directory holds several timestamped sample files (re-runs or merged retries), the latest one is used for the reports. 同一任务有多个时间戳文件时，报表取最新一份。
- Endpoints and API keys are redacted. 端点与 API key 已脱敏。
- Scores from different calibers (full A/B tier vs. class-balanced C tier, LLM-judged, textualized) are not
comparable; see
reports/README_capability.md.
Models: qwen3.8-max-0902, qwen3.8-27b, qwen3.8-flash, doubao-seed-2-1-pro-260915, gemma-4-26b-a4b-it, mimo-v2.6-flash, mimo-v2.6-pro, minimax-m3, glm-5.3-flash, deepseek-v4.1-flash.
- Downloads last month
- 51