Skip to main content
A model family for predicting the next item from an ordered interaction history. The assigned model card identifies the actual architecture and training lineage. Context length and embedding size are properties of the deployed artifact, not fixed family promises. Requests exceeding its limit are rejected instead of silently truncated.

​
Evaluation

Research has measured cross-category next-item transfer on selected review and rating datasets. It has not established universal transfer, action-type prediction, or causal simulation. Your deployment’s report specifies its dataset, temporal split, baselines, and sample size. See Models, Predict, and Adaptation.