Datasetpublic

egygi/computer-use-large-actions

computer-use-large-actions 9,000 instruction pairs derived from markov-ai/computer-use-large descriptions.parquet (not the raw 12,300 hours of video). Each row is a 10-second segment whose LLM description is a real GUI action (NO_TASK dropped). Source license is CC-BY-4.0. Split by software category examples vscode 2,500 autocad 2,500 blender 1,000 excel 1,000 photoshop 1,000 salesforce 1,000 VS Code and AutoCAD are oversampled for… See the full description on the dataset page: https://huggingface.co/datasets/egygi/computer-use-large-actions.

sourceHugging Facecc-by-4.0updated 23d agoView on Hugging Face

0likes52downloads

Dataset Card

computer-use-large-actions

9,000 instruction pairs derived from `markov-ai/computer-use-large` descriptions.parquet (not the raw 12,300 hours of video).

Each row is a 10-second segment whose LLM description is a real GUI action (NO_TASK dropped). Source license is CC-BY-4.0.

Split by software

category	examples
vscode	2,500
autocad	2,500
blender	1,000
excel	1,000
photoshop	1,000
salesforce	1,000

VS Code and AutoCAD are oversampled for Emmanuel's agentic-coding and DigiCatWalk CAD work.

Format

messages for TRL SFT: system + user (software / tutorial title / ASR transcript) → assistant (one-sentence GUI action).

What this is not

This is not a click/xy action trace and not a video-VLM dataset. The videos stay on markov-ai/computer-use-large. Use those later to add frames at t0–t1 for a vision pass.