Use case · Idea to test
Product recommendations
Rank available products by personal taste.
- Rate a few products
- Filter price and stock
- Rank the remaining choices
A shopper could teach a catalog their taste in furniture or household items, then browse a short list within a chosen budget.
Where the small model fits
A product model represents the existing inventory. The small preference scorer ranks real items; no new product has to be generated. Price, availability and required features are separate filters.
What would need to work
Compare choices against ordinary filters and recommendations, including returns and variety. Preference should be measured on unseen products rather than the examples used to train the scorer.
Try the related work
This demo learns from simple drawings. It does not run the application described above.