SCREENCOMPUTER USE
Software Task Screenflows
Desktop task recordings that connect interface state, user action and task outcome across a continuous workflow.
View datasetMULTIMODALPAIRED CAPTURE
Multimodal Paired Capture Sets
Multiple modalities are only useful together when their relationship is clear. Multimodal Paired Capture Sets are designed around that correspondence, linking selected audio, video, screen, image or text records through shared identifiers and timing information.
The pairing can exist at file, session, segment or timestamp level depending on the task. That makes the cross-modal relationship part of the supervision itself, rather than leaving a technical team to infer how separately delivered media streams line up.
Technical structure
- Shared identifiers connecting the selected modalities
- Timestamp or segment alignment at the level required by the task
- Cross-modal manifests describing record relationships
- Task-specific examples linking source media to the relevant paired signal
Typical applications
Multimodal representation learning, vision-language models, audio-visual modelling, cross-modal retrieval and multimodal evaluation.
Delivery note
The modality set and alignment precision are defined according to the relationship the model needs to learn.
Delivery
The exact package depends on the collection scope. Where relevant, delivery can include task-specific records, manifests, provenance fields, stable identifiers, SHA-256 hashes, MLCommons Croissant 1.0 metadata and loading instructions.