Search papers, labs, and topics across Lattice.
This paper addresses the challenge of enhancing LLM-based generative recommendation models by integrating personalized natural language to restore collaborative signals in the generation of semantic-ID (SID) tuples. The authors identify that traditional methods struggle due to the competition between content and collaborative signals within the compact SID representation, leading to poor recommendations. By introducing a framework that utilizes hierarchical collaborative cues during SID generation, the study achieves consistent improvements in recommendation accuracy without the need for retraining or altering the model's backbone.
Restoring collaborative signals in generative recommendations can significantly boost accuracy without retraining models or sacrificing efficiency.
Making LLM-based generative recommendation models stronger and more personalized through natural language and explicit reasoning is a widely anticipated yet still unsolved goal. Such models cast recommendation as autoregressively generating an item's semantic-ID (SID), a short tuple of discrete codes, so that recommending well reduces to emitting the right SID. In this setting the model verbalizes its knowledge poorly, and text and SID tokens live in misaligned embedding spaces. Deep reasoning therefore rarely turns into a correct SID, and enabling explicit"thinking"often gives no gain or even hurts. The deeper cause is that a compact SID cannot hold content and collaborative signal at once: the two compete, and collaboration loses. Because a mis-predicted SID is a wrong recommendation, this caps accuracy directly. Costly multi-round training barely helps, and few methods try to supply the missing signal at inference time. What is missing is a reliable channel that carries collaborative signal into SID generation. We therefore propose a framework, guided by personalized natural language, that adds hierarchical collaborative cues as the model generates, without altering the backbone or retraining the SIDs. Rather than mapping language onto SIDs directly, it uses language to attach analyzable links between collaborative patterns and their audiences, restoring the collaborative signal that SIDs miss. The result is consistent gains in recommendation accuracy, grounding generation in collaborative structure at inference time rather than relying on explicit reasoning or retraining.
