Rollouts
Collection
5 items • Updated
Datasets:
images images list | messages string | metadata string |
|---|---|---|
[{"role": "user", "content": [{"type": "image", "index": 0.0, "text": null}, {"type": "text", "index": null, "text": "You are preparing a session for an Automated Dialogue Replacement (ADR) recording where an actor needs to re-record a line from 5.0 to 15.0 seconds. Configure the 'MyProject' Ardour session by creating ... | {"platform": "desktop", "task_type": "use", "extra_tool_schemas": [], "valid_actions": null, "others": {"source": "gym-anything/ardour_env", "upstream_task_id": "adr_session_prep@1", "upstream_env_id": "ardour_env@0.1", "reward_type": "sparse", "env_id": "lite.cuaworld.ardour", "task_id": "adr_session_prep", "model_id"... | |
[{"role": "user", "content": [{"type": "image", "index": 0.0, "text": null}, {"type": "text", "index": null, "text": "You are an audio archivist at a historical society processing a newly digitized oral history tape from /home/ga/Audio/samples/oral_history_raw.wav. In Ardour, create a track named 'Oral History 1974' an... | {"platform": "desktop", "task_type": "use", "extra_tool_schemas": [], "valid_actions": null, "others": {"source": "gym-anything/ardour_env", "upstream_task_id": "archival_audio_restoration@1", "upstream_env_id": "ardour_env@0.1", "reward_type": "sparse", "env_id": "lite.cuaworld.ardour", "task_id": "archival_audio_rest... | |
[{"role": "user", "content": [{"type": "image", "index": 0.0, "text": null}, {"type": "text", "index": null, "text": "You are an audio engineer creating content for a wellness app. Using the mono voice recording at /home/ga/Audio/client_files/narration.wav, create an immersive 'ear-to-ear' ASMR whisper track using the ... | {"platform": "desktop", "task_type": "use", "extra_tool_schemas": [], "valid_actions": null, "others": {"source": "gym-anything/ardour_env", "upstream_task_id": "asmr_haas_stereo_widening@1", "upstream_env_id": "ardour_env@0.1", "reward_type": "sparse", "env_id": "lite.cuaworld.ardour", "task_id": "asmr_haas_stereo_wid... | |
[{"role": "user", "content": [{"type": "image", "index": 0.0, "text": null}, {"type": "text", "index": null, "text": "You are a freelance audiobook narrator working with Blackthorn Publishing. You have just finished recording Chapter 5 of the novel 'The Last Meridian' by Eleanor Voss and need to prepare the recording f... | {"platform": "desktop", "task_type": "use", "extra_tool_schemas": [], "valid_actions": null, "others": {"source": "gym-anything/ardour_env", "upstream_task_id": "audiobook_chapter_export@1", "upstream_env_id": "ardour_env@0.1", "reward_type": "sparse", "env_id": "lite.cuaworld.ardour", "task_id": "audiobook_chapter_exp... | |
[{"role": "user", "content": [{"type": "image", "index": 0.0, "text": null}, {"type": "text", "index": null, "text": "You are a wildlife biologist extracting rare bird calls from a continuous field recording. Based on the field notes at /home/ga/Audio/field_notes.txt, extract the three specified timestamp ranges from t... | {"platform": "desktop", "task_type": "use", "extra_tool_schemas": [], "valid_actions": null, "others": {"source": "gym-anything/ardour_env", "upstream_task_id": "bioacoustics_extraction@1", "upstream_env_id": "ardour_env@0.1", "reward_type": "sparse", "env_id": "lite.cuaworld.ardour", "task_id": "bioacoustics_extractio... | |
[{"role": "user", "content": [{"type": "image", "index": 0.0, "text": null}, {"type": "text", "index": null, "text": "You are a broadcast audio technician preparing a syndicated radio interview for daytime airing. Replace the expletive located between 14.5 and 15.5 seconds in the raw interview with the 1kHz censor beep... | {"platform": "desktop", "task_type": "use", "extra_tool_schemas": [], "valid_actions": null, "others": {"source": "gym-anything/ardour_env", "upstream_task_id": "broadcast_censor_edit@1", "upstream_env_id": "ardour_env@0.1", "reward_type": "sparse", "env_id": "lite.cuaworld.ardour", "task_id": "broadcast_censor_edit", ... | |
[{"src":"https://huggingface.co/proxy/datasets-server.huggingface.co/assets/cua-lite/Lite.CUAWorld/--/{dataset_git_revisi(...TRUNCATED) | "[{\"role\": \"user\", \"content\": [{\"type\": \"image\", \"index\": 0.0, \"text\": null}, {\"type\(...TRUNCATED) | "{\"platform\": \"desktop\", \"task_type\": \"use\", \"extra_tool_schemas\": [], \"valid_actions\": (...TRUNCATED) |
[{"src":"https://huggingface.co/proxy/datasets-server.huggingface.co/assets/cua-lite/Lite.CUAWorld/--/{dataset_git_revisi(...TRUNCATED) | "[{\"role\": \"user\", \"content\": [{\"type\": \"image\", \"index\": 0.0, \"text\": null}, {\"type\(...TRUNCATED) | "{\"platform\": \"desktop\", \"task_type\": \"use\", \"extra_tool_schemas\": [], \"valid_actions\": (...TRUNCATED) |
[{"src":"https://huggingface.co/proxy/datasets-server.huggingface.co/assets/cua-lite/Lite.CUAWorld/--/{dataset_git_revisi(...TRUNCATED) | "[{\"role\": \"user\", \"content\": [{\"type\": \"image\", \"index\": 0.0, \"text\": null}, {\"type\(...TRUNCATED) | "{\"platform\": \"desktop\", \"task_type\": \"use\", \"extra_tool_schemas\": [], \"valid_actions\": (...TRUNCATED) |
[{"src":"https://huggingface.co/proxy/datasets-server.huggingface.co/assets/cua-lite/Lite.CUAWorld/--/{dataset_git_revisi(...TRUNCATED) | "[{\"role\": \"user\", \"content\": [{\"type\": \"image\", \"index\": 0.0, \"text\": null}, {\"type\(...TRUNCATED) | "{\"platform\": \"desktop\", \"task_type\": \"use\", \"extra_tool_schemas\": [], \"valid_actions\": (...TRUNCATED) |
End of preview. Expand in Data Studio
cua-lite/Lite.CUAWorld
Lite.CUAWorld GPT-5.5 grounded teacher trajectories (train, 38 gym-anything software desktops); every trajectory kept, quality gates tagged in metadata.others.exclude_reason (filter with not exclude_reason and episode_return>0.5)
Origin
Load via datasets
from datasets import load_dataset
# entire dataset
ds = load_dataset("cua-lite/Lite.CUAWorld")
# just one named subset (config)
ds = load_dataset("cua-lite/Lite.CUAWorld", "desktop.use.ardour")
You can also filter by metadata.platform / metadata.task_type /
metadata.others.* after loading; every row carries a rich metadata
struct (see schema below).
Schema
Each row has these columns:
| column | type | notes |
|---|---|---|
images |
list[Image] | embedded PNG/JPEG bytes; HF viewer renders thumbnails |
messages |
list[struct] | OpenAI-style turns with role + structured content |
metadata |
struct | {platform, task_type, extra_tool_schemas, valid_actions, others{...}} |
Coordinate values in messages are normalized to [0, 1000] integers.
Image-dedup (grounding.* / understanding cohorts). These cohorts are
single-image-per-row and many rows share the same screenshot, so to avoid
re-embedding identical image bytes once per instruction they are stored
folded: one row per unique screenshot (image embedded once), carrying an
extra _folded column — a JSON string with the authoritative list of
{messages, metadata} members for that screenshot. The row's top-level
messages is the members concatenated for viewer convenience. use
cohorts are not folded. Use lite.data.hf.download to consume this repo —
it unfolds automatically back to one row per instruction; reading the parquet
directly yields the folded form.
Layout
<platform>/<task_type>/<split>/shard-NNNNN-of-NNNNN.parquet # single-variant cohort
<platform>/<task_type>/<split>/<variant>/shard-NNNNN-of-NNNNN.parquet # multi-variant cohort
platform∈ {desktop, mobile, web}task_type∈ {understanding, grounding.action, grounding.point, grounding.bbox, use} — used verbatim as the dir component- HF config names are
<platform>.<task_type>by default (e.g.mobile.grounding.action) — UNLESS the dataset was staged with--config-names, which sets verbatim, explicitly-chosen config names (see theconfigs:block above for the authoritative list). The agent registry lookup key in code is<agent>@<platform>@<task_type>(e.g.qwen3_vl@mobile@grounding.action); only this user-facing token uses.between platform and task_type, because@triggers a 403 on the dataset-viewer's signed image URLs. - HF split names stay
train/validation(thedatasetslibrary blacklists<>:/\|?*in split names; everything else is fine in config_name) validationis an in-distribution held-out slice (never used in training);testis reserved for out-of-distribution benchmark datasets
Stats
| platform | task_type | variant | train | validation |
|---|---|---|---|---|
| desktop | use | desktop.use.ardour | 32 | 0 |
| desktop | use | desktop.use.astroimagej | 35 | 0 |
| desktop | use | desktop.use.blender3d | 43 | 0 |
| desktop | use | desktop.use.coppeliasim | 27 | 0 |
| desktop | use | desktop.use.dbeaver | 53 | 0 |
| desktop | use | desktop.use.diagrams_net | 45 | 0 |
| desktop | use | desktop.use.eclipse | 30 | 0 |
| desktop | use | desktop.use.gcompris | 39 | 0 |
| desktop | use | desktop.use.geogebra | 57 | 0 |
| desktop | use | desktop.use.gmat | 43 | 0 |
| desktop | use | desktop.use.gpredict | 31 | 0 |
| desktop | use | desktop.use.gretl | 46 | 0 |
| desktop | use | desktop.use.gvsig_desktop | 50 | 0 |
| desktop | use | desktop.use.hec_ras | 48 | 0 |
| desktop | use | desktop.use.imagej | 55 | 0 |
| desktop | use | desktop.use.jstock | 27 | 0 |
| desktop | use | desktop.use.kstars_sim | 51 | 0 |
| desktop | use | desktop.use.librecad | 54 | 0 |
| desktop | use | desktop.use.libreoffice_calc | 72 | 0 |
| desktop | use | desktop.use.moodle | 54 | 0 |
| desktop | use | desktop.use.odoo | 37 | 0 |
| desktop | use | desktop.use.openemr | 55 | 0 |
| desktop | use | desktop.use.openlca | 51 | 0 |
| desktop | use | desktop.use.openrocket | 53 | 0 |
| desktop | use | desktop.use.openvsp | 48 | 0 |
| desktop | use | desktop.use.pycharm | 29 | 0 |
| desktop | use | desktop.use.pymol | 54 | 0 |
| desktop | use | desktop.use.qblade | 47 | 0 |
| desktop | use | desktop.use.qgis | 60 | 0 |
| desktop | use | desktop.use.slicer3d | 143 | 0 |
| desktop | use | desktop.use.solvespace | 54 | 0 |
| desktop | use | desktop.use.sumo | 42 | 0 |
| desktop | use | desktop.use.sweet_home_3d | 55 | 0 |
| desktop | use | desktop.use.ugene | 43 | 0 |
| desktop | use | desktop.use.vlc_media_player | 38 | 0 |
| desktop | use | desktop.use.vscode | 37 | 0 |
| desktop | use | desktop.use.webots | 50 | 0 |
| desktop | use | desktop.use.wordpress | 28 | 0 |
Local mirror & SFT export
For local workflows (SFT export, dedup, mixing across datasets), use
lite.data.hf.download to mirror this repo back to the canonical local
layout:
$CUA_LITE_DATASETS_ROOT/cua-lite/Lite.CUAWorld/
images/<hash[:2]>/<hash>.<ext> # content-addressed image store
<platform>/<task_type>/<split>[/<variant>].parquet # rows reference images by relative path
Rows in the local parquet have images: list[str]; bytes are extracted to
the image store. lite.train.export.export_sft consumes the local
form directly with --image-root=$CUA_LITE_DATASETS_ROOT.
- Total unique images: 34,478
- Image store size: 18.89 GB
Notes
Staged via lite.data.hf.stage from rollout log-roots: .data/rollout/lite.cuaworld/gpt/b236be6d/ardour/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/astroimagej/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/blender3d/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/coppeliasim/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/dbeaver/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/diagrams_net/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/eclipse/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/gcompris/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/geogebra/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/gmat/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/gpredict/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/gretl/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/gvsig_desktop/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/hec_ras/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/imagej/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/jstock/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/kstars_sim/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/librecad/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/libreoffice_calc/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/moodle/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/odoo/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/openemr/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/openlca/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/openrocket/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/openvsp/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/pycharm/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/pymol/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/qblade/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/qgis/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/slicer3d/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/solvespace/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/sumo/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/sweet_home_3d/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/ugene/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/vlc_media_player/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/vscode/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/webots/train_annotated, .data/rollout/lite.cuaworld/gpt/b236be6d/wordpress/train_annotated (row filter: none).
License & citation
other
- Downloads last month
- 231