SenseRequest a dataset

Human data from the people
who understand people.

Demonstrations, preferences, and evals from professional creatives. Text, audio, video. Consented, provenance on every sample.

01

Demonstrations

Long-horizon, in character, on your prompts.

02

Preferences

Ranked and rationaled. Reward-model ready.

03

Evals

Held-out, human-graded, where the model reads a person wrong.

04

Judgment

The turn where the model should step out of the role.

Creatives, not crowdworkers

Writers, actors, and directors. Reading people is the job.

Your format

SFT, pairwise, rubric-scored. JSONL, audio, video. Delivered the way your pipeline reads it.

Consent, built in

Every contribution scoped, paid, and revocable. Provenance on every sample.

Ready to build better AI? Get in touch to discuss your data needs and explore how professional creatives are transforming AI.

Request a dataset