New research tackles LLM integration for generative recommendation systems · 8 sources tracked
ByPulseAugur Editorial·[16 sources]·
Several new research papers explore advancements in generative recommendation systems, focusing on how to better integrate user behavior and item semantics into large language models (LLMs). G2Rec proposes a scalable framework that unifies graph-based user co-engagement modeling with semantic tokenization. Token Factory offers a method to transform traditional signals into 'soft tokens' for efficient integration into LLMs, preventing prompt length issues. Another study investigates LLMs' memorization behavior in generative recommendation and introduces IIRG to capture richer item-item relations. HoloRec presents an endogenous chain-of-thought mechanism for unified representation, reasoning, and generation. Additionally, research examines whether generative recommenders deepen information cocoons, finding they are generally less prone than traditional methods but can still exhibit concentration within generated semantic spaces, influenced by tokenization strategy and model scale. Finally, ReaEmb harmonizes semantic and collaborative signals in LLMs for sequential recommendation by enhancing reasoning and explicitly injecting collaborative signals.
AI
IMPACT
These papers introduce novel techniques for integrating user data and item semantics into LLMs for recommendation, potentially improving personalization and addressing issues like memorization and information cocoons.
RANK_REASON
Cluster consists of multiple academic papers published on arXiv detailing new methods and analyses in generative recommendation systems.
arXiv:2606.20554v1 Announce Type: cross Abstract: Generative recommendation is an emerging paradigm that has shown promise in industrial recommendation systems, aiming to predict users' next interactions from their historical behaviors. At the core of generative recommendation li…
arXiv:2606.19635v1 Announce Type: cross Abstract: Large Recommendation Models (LRMs) have demonstrated promising capabilities in industry-scale recommendation tasks. However, holistically integrating traditional signals into these transformer-based architectures effectively and e…
Generative recommendation is an emerging paradigm that has shown promise in industrial recommendation systems, aiming to predict users' next interactions from their historical behaviors. At the core of generative recommendation lies item tokenization, which bridges item semantics…
Large Recommendation Models (LRMs) have demonstrated promising capabilities in industry-scale recommendation tasks. However, holistically integrating traditional signals into these transformer-based architectures effectively and efficiently remains a major challenge. Conventional…
arXiv cs.LG
TIER_1English(EN)·Sunwoo Kim, Sunkyung Lee, Clark Mingxuan Ju, Donald Loveland, Bhuvesh Kumar, Kijung Shin, Neil Shah, Liam Collins·
arXiv:2606.17276v1 Announce Type: cross Abstract: Generative recommendation (GR) has emerged as a promising direction for recommender systems. Recently, large language models (LLMs) have been increasingly adopted for GR, as their rich pretrained knowledge is expected to help them…
Recommender systems alleviate information overload, yet repeated feedback between recommendations and user interactions can reinforce existing preferences and narrow users' exposure, forming information cocoons. While this phenomenon has been widely studied in traditional sequent…
arXiv:2606.15331v1 Announce Type: cross Abstract: Generative recommendation models that formulate the task as sequence generation overcome the objective fragmentation problem of traditional cascade architectures, yet existing approaches still suffer from flat semantic representat…
Generative recommendation (GR) has emerged as a promising direction for recommender systems. Recently, large language models (LLMs) have been increasingly adopted for GR, as their rich pretrained knowledge is expected to help them generalize beyond common user behavior patterns t…
Generative recommendation (GR) has emerged as a promising direction for recommender systems. Recently, large language models (LLMs) have been increasingly adopted for GR, as their rich pretrained knowledge is expected to help them generalize beyond common user behavior patterns t…
Sequential Recommender Systems (SRS) predict the next item of interest based on users' interaction histories and have been widely deployed, but hindered by long-tail problem. Large Language Models (LLMs), with strong semantic understanding and reasoning capabilities, offer a prom…
arXiv cs.AI
TIER_1English(EN)·Yinhan He, Liam Collins, Bhuvesh Kumar, Jundong Li, Neil Shah, Donald Loveland·
arXiv:2606.14142v1 Announce Type: cross Abstract: Large Language Models (LLMs) are increasingly adopted as backbones for Generative Recommendation (GR), promising access to pretrained world knowledge. Yet reliably invoking this knowledge for GR remains poorly understood. A key ob…
arXiv:2605.17779v2 Announce Type: replace Abstract: Generative recommendation reformulates recommendation as next-token prediction over discrete semantic identifiers (IDs). A fundamental yet unexplored design choice is that existing methods employ fixed-length tokenization for al…
Large Language Models for generative recommendation face challenges with semantic IDs disrupting natural-language reasoning, prompting a lightweight implicit reasoning approach that outperforms explicit methods while reducing computational costs.
Existing sequential recommendation models rely on dataset-specific training, where the learned parameters are fitted to the item catalog and the observed interaction distribution of the training data. This limits generalization to new domains, typically requiring retraining from …
Generative recommendation models that formulate the task as sequence generation overcome the objective fragmentation problem of traditional cascade architectures, yet existing approaches still suffer from flat semantic representations lacking hierarchical structure for multi-step…
Large Language Models (LLMs) are increasingly adopted as backbones for Generative Recommendation (GR), promising access to pretrained world knowledge. Yet reliably invoking this knowledge for GR remains poorly understood. A key obstacle is that LLM-based GR typically represents i…
