Datasets:
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
Error code: DatasetGenerationError
Exception: ArrowInvalid
Message: Schema at index 1 was different:
checkpoint: string
num_sequences: int64
ep_len: int64
eval_mode: string
eval_scan_sequences: int64
eval_sequence_offset: int64
g_scale: double
query_g_control: string
online_minimal_g_mode: string
behavior_contact_radius: double
behavior_tool_clearance: double
behavior_contact_tolerance: double
behavior_motion_threshold: double
language_control: string
action_execution_horizon: int64
temporal_ensemble: bool
temporal_ensemble_decay: double
state_only_observation: bool
policy_sample_seed: null
fm_reseed_each_episode: bool
trace_rollouts: int64
selected_tasks: list<item: string>
selected_sequence_indices: list<item: int64>
raw_results: list<item: int64>
vs
eval_mode: string
g_scale: double
query_g_control: string
online_minimal_g_mode: string
language_control: string
action_execution_horizon: int64
temporal_ensemble: bool
temporal_ensemble_decay: double
state_only_observation: bool
policy_sample_seed: null
fm_reseed_each_episode: bool
avg_seq_len: double
chain_sr: struct<1: double>
raw_results: list<item: int64>
task_info: struct<push_blue_block_left: struct<success: int64, total: int64, sr: double>, push_red_block_right: struct<success: int64, total: int64, sr: double>>
eval_sequence_offset: int64
selected_tasks: list<item: string>
selected_sequence_indices: list<item: int64>
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1856, in _prepare_split_single
writer.write_table(table)
~~~~~~~~~~~~~~~~~~^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 810, in write_table
self.write_rows_on_file() # in case there are buffered rows to write first
~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 689, in write_rows_on_file
table = pa.concat_tables(self.current_rows)
File "pyarrow/table.pxi", line 6268, in pyarrow.lib.concat_tables
File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
return check_status(status)
File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
raise convert_status(status)
pyarrow.lib.ArrowInvalid: Schema at index 1 was different:
checkpoint: string
num_sequences: int64
ep_len: int64
eval_mode: string
eval_scan_sequences: int64
eval_sequence_offset: int64
g_scale: double
query_g_control: string
online_minimal_g_mode: string
behavior_contact_radius: double
behavior_tool_clearance: double
behavior_contact_tolerance: double
behavior_motion_threshold: double
language_control: string
action_execution_horizon: int64
temporal_ensemble: bool
temporal_ensemble_decay: double
state_only_observation: bool
policy_sample_seed: null
fm_reseed_each_episode: bool
trace_rollouts: int64
selected_tasks: list<item: string>
selected_sequence_indices: list<item: int64>
raw_results: list<item: int64>
vs
eval_mode: string
g_scale: double
query_g_control: string
online_minimal_g_mode: string
language_control: string
action_execution_horizon: int64
temporal_ensemble: bool
temporal_ensemble_decay: double
state_only_observation: bool
policy_sample_seed: null
fm_reseed_each_episode: bool
avg_seq_len: double
chain_sr: struct<1: double>
raw_results: list<item: int64>
task_info: struct<push_blue_block_left: struct<success: int64, total: int64, sr: double>, push_red_block_right: struct<success: int64, total: int64, sr: double>>
eval_sequence_offset: int64
selected_tasks: list<item: string>
selected_sequence_indices: list<item: int64>
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1876, in _prepare_split_single
num_examples, num_bytes = writer.finalize()
~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 830, in finalize
self.write_rows_on_file()
~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 689, in write_rows_on_file
table = pa.concat_tables(self.current_rows)
File "pyarrow/table.pxi", line 6268, in pyarrow.lib.concat_tables
File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
raise convert_status(status)
pyarrow.lib.ArrowInvalid: Schema at index 1 was different:
checkpoint: string
num_sequences: int64
ep_len: int64
eval_mode: string
eval_scan_sequences: int64
eval_sequence_offset: int64
g_scale: double
query_g_control: string
online_minimal_g_mode: string
behavior_contact_radius: double
behavior_tool_clearance: double
behavior_contact_tolerance: double
behavior_motion_threshold: double
language_control: string
action_execution_horizon: int64
temporal_ensemble: bool
temporal_ensemble_decay: double
state_only_observation: bool
policy_sample_seed: null
fm_reseed_each_episode: bool
trace_rollouts: int64
selected_tasks: list<item: string>
selected_sequence_indices: list<item: int64>
raw_results: list<item: int64>
vs
eval_mode: string
g_scale: double
query_g_control: string
online_minimal_g_mode: string
language_control: string
action_execution_horizon: int64
temporal_ensemble: bool
temporal_ensemble_decay: double
state_only_observation: bool
policy_sample_seed: null
fm_reseed_each_episode: bool
avg_seq_len: double
chain_sr: struct<1: double>
raw_results: list<item: int64>
task_info: struct<push_blue_block_left: struct<success: int64, total: int64, sr: double>, push_red_block_right: struct<success: int64, total: int64, sr: double>>
eval_sequence_offset: int64
selected_tasks: list<item: string>
selected_sequence_indices: list<item: int64>
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1379, 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 956, 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 1702, 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 1887, in _prepare_split_single
raise DatasetGenerationError("An error occurred while generating the dataset") from e
datasets.exceptions.DatasetGenerationError: An error occurred while generating the datasetNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
rollout_index int64 | sequence_index int64 | task string | success bool | steps list |
|---|---|---|---|---|
0 | 87 | push_blue_block_left | false | [
{
"step": 0,
"eef": [
0.0258684251,
-0.2313131839,
0.5712804794
],
"action": [
0.053150136,
0.3490435779,
0.2698428333,
-0.0191293191,
-0.2081958801,
0.1974499822,
1
],
"block_center": [
0.0499557368,
-0.1199908704,
... |
1 | 100 | push_blue_block_left | false | [
{
"step": 0,
"eef": [
0.0258684251,
-0.2313131839,
0.5712804794
],
"action": [
-0.0929021463,
0.2653211355,
-0.045151256,
-0.0818892941,
0.0138028804,
0.1939023882,
-1
],
"block_center": [
0.049999848,
-0.1200000122,
... |
2 | 139 | push_blue_block_left | false | [{"step":0,"eef":[0.0258684251,-0.2313131839,0.5712804794],"action":[-0.0723856166,0.2648710907,-0.0(...TRUNCATED) |
3 | 195 | push_blue_block_left | false | [{"step":0,"eef":[0.0258684251,-0.2313131839,0.5712804794],"action":[-0.079503946,0.2564711571,-0.05(...TRUNCATED) |
4 | 221 | push_blue_block_left | true | [{"step":0,"eef":[0.0258684251,-0.2313131839,0.5712804794],"action":[1.0,-0.094873473,0.2154357135,-(...TRUNCATED) |
5 | 244 | push_blue_block_left | false | [{"step":0,"eef":[0.0258684251,-0.2313131839,0.5712804794],"action":[-0.0162716694,0.2565408945,-0.0(...TRUNCATED) |
6 | 280 | push_blue_block_left | true | [{"step":0,"eef":[0.0258684251,-0.2313131839,0.5712804794],"action":[0.059149269,0.2373142689,-0.012(...TRUNCATED) |
7 | 315 | push_blue_block_left | true | [{"step":0,"eef":[0.0258684251,-0.2313131839,0.5712804794],"action":[1.0,-0.0797383636,0.2148143053,(...TRUNCATED) |
8 | 335 | push_blue_block_left | true | [{"step":0,"eef":[0.0258684251,-0.2313131839,0.5712804794],"action":[1.0,-0.0962915719,0.2157692611,(...TRUNCATED) |
9 | 349 | push_blue_block_left | false | [{"step":0,"eef":[0.0258684251,-0.2313131839,0.5712804794],"action":[-0.0686252639,0.2501865625,-0.0(...TRUNCATED) |
Gina AutoResearch — derived multi-simulator trajectory release
This release contains project-produced semantic indexes, intent/phase labels, exact upstream acquisition manifests, and five newly executed RoboTwin expert trajectories. It deliberately does not mirror approximately 90 GB of official CALVIN, LIBERO, RoboTwin, ManiSkill, or VIMA payloads.
Code and reconstruction scripts: https://github.com/Austingina/Gina_autoresearch
Pinned code revision for this dataset: 3fe4db288a63ddef3e08818884a30fdc2d7e17d9 (also recorded in CODE_REVISION.json).
Scope
| Source | Successful parents | Tasks | Native transitions/frames | Payload policy |
|---|---|---|---|---|
| CALVIN | 432 | 12 | 25,852 | exact source URL and segment intervals |
| LIBERO | 2,000 | 40 | 338,575 | pinned HF revision, paths and SHA-256 |
| RoboTwin official | 2,500 | 50 | 548,893 | pinned HF revision, archive paths and hashes |
| ManiSkill | 15,262 | 9 | 1,581,768 | pinned HF revision, 9 archive paths and SHA-256 |
| VIMA-Bench | 5,999 | 6 | 12,668 actions | pinned HF revision plus member paths, CRC32 and SHA-256 |
| RoboTwin online-native pilot | 5 | 1 | 866 | five HDF5 payloads included |
The canonical official view is semantic_indexes/official_reorg_20261023_v6/: 26,193 unique successful parent trajectories, 117 simulator-qualified tasks, 2,507,756 native transitions/frames, and 95,033 derived phase views. Phase views are labels over parent demonstrations and are not independent success evidence.
Guidance field semantics
The compact candidate field is G=[A,F], where A is a 3-D vector from the current end effector to a task-conditioned interaction anchor and F is a 3-D desired object-motion direction. G is an external, recomputable sidecar; it is not an admission gate, not a simulator action, and not enabled as a default model input in this release.
Reconstruction
- Clone the GitHub repository.
- Read
source_manifests/<simulator>/for the exact source repository/URL, pinned revision, native path or archive member, size, CRC32 and/or SHA-256. - Run the corresponding scripts under
framework/scripts/data/orscripts/. - Verify downloaded bytes against the recorded hashes and rebuild the canonical index.
Absolute local paths retained inside historical audit records are provenance strings from the source machine; consumers should resolve payloads using the relative paths and pinned source manifests.
Splits and evaluation status
CALVIN's 12-fold manifest records 324 training segments and 27 held-out segments per fold. The published multi-source data is a semantic inventory, not a single merged action/state tensor and not a claim of cross-simulator policy success. CALVIN has closed-loop G evaluations; the other sources in this release have passed offline native-format/readback validation, while cross-simulator six-dimensional-G policy training and native-Oracle rollouts remain incomplete.
Licensing
This repository contains mixed-source metadata. Original payloads remain governed by their upstream licenses and terms; consult each pinned source before downloading. Project-generated indexes and labels are provided for research/audit use, and the five included RoboTwin pilot HDF5 files should be used consistently with RoboTwin's upstream terms. No upstream license is overridden by this dataset card.
2026-10-26 guidance / C-K / phase supplement
See supplements/2026-10-26_guidance_ck_p/README.md. This additive release pins GitHub commit 671764be7161aa51261898c322c55f20d3a5d175 and preserves compact sidecars, selected checkpoints, metrics, native rollout evidence, and CALVIN recovery receipts. It does not publish large third-party official payloads or claim the incomplete CALVIN mechanism cohort is admitted.
- Downloads last month
- 142