arXiv:2510.21276v1 Announce Type: cross 
Abstract: Generative recommendation (GR) models tokenize each action into a few discrete tokens (called semantic IDs) and autoregressively generate the next tokens as predictions, showing advantages such as memory efficiency, scalability, and the potential to unify retrieval and ranking. Despite these benefits, existing tokenization methods are static and non-personalized. They typically derive semantic IDs solely from item features, assuming a universal item similarity that overlooks user-specific perspectives. However, under the autoregressive paradigm, semantic IDs with the same prefixes always receive similar probabilities, so a single fixed mapping implicitly enforces a universal item similarity standard across all users. In practice, the same item may be interpreted differently depending on user intentions and preferences. To address this issue, we propose a personalized context-aware tokenizer that incorporates a user's historical interactions when generating semantic IDs. This design allows the same item to be tokenized into different semantic IDs under different user contexts, enabling GR models to capture multiple interpretive standards and produce more personalized predictions. Experiments on three public datasets demonstrate up to 11.44% improvement in NDCG@10 over non-personalized action tokenization baselines. Our code is available at https://github.com/YoungZ365/Pctx.

تقدم الورقة الأخيرة حول Pctx نهجًا مبتكرًا لنماذج التوصية التوليدية من خلال تحويل السياق الشخصي إلى رموز، مما يعزز كفاءة ومرونة التنبؤات. هذه الابتكار مهم لأنه يعالج قيود طرق التحويل التقليدية التي تفتقر إلى التخصيص، مما قد يحول الطريقة التي يتم بها إنشاء التوصيات ويحسن تجارب المستخدمين عبر منصات مختلفة.

El reciente artículo sobre Pctx presenta un enfoque innovador para los modelos de recomendación generativa al tokenizar el contexto personalizado, lo que mejora la eficiencia y escalabilidad de las predicciones. Esta innovación es significativa porque aborda las limitaciones de los métodos de tokenización tradicionales que carecen de personalización, lo que podría transformar la forma en que se generan las recomendaciones y mejorar la experiencia del usuario en diversas plataformas.

Le récent article sur Pctx présente une approche révolutionnaire des modèles de recommandation générative en tokenisant le contexte personnalisé, ce qui améliore l'efficacité et l'évolutivité des prédictions. Cette innovation est importante car elle répond aux limites des méthodes de tokenisation traditionnelles qui manquent de personnalisation, transformant potentiellement la manière dont les recommandations sont générées et améliorant l'expérience utilisateur sur diverses plateformes.

The recent paper on Pctx introduces a groundbreaking approach to generative recommendation models by tokenizing personalized context, which enhances the efficiency and scalability of predictions. This innovation is significant because it addresses the limitations of traditional tokenization methods that lack personalization, potentially transforming how recommendations are generated and improving user experiences across various platforms.

Pctx: Tokenizing Personalized Context for Generative Recommendation

One More Thing in AI – Your Shortcut to AI Mastery

Pctx: Tokenizing Personalized Context for Generative Recommendation

Was this article worth reading? Share it

One More Thing in AI

Airparser

CRANQ

Scop.ai

ProGiftIdeas

Typewell

Ready to build your own newsroom?