TSPORec: Token Selection via Preference Optimization for LLM-Based Sequential Recommendation

RosePO: Customized Preference Alignment in LLM-Based Recommendation

Recent advancements in Large Language Models (LLMs) have inspired their application in sequential recommendation systems, often through supervised fine-tuning (SFT). However, conventional SFT methods often struggle to capture nuanced comparative relationships between items. While recent approaches utilize Direct Preference Optimization (DPO), they remain constrained by challenges in modeling customized preferences, including capturing fine-grained user preferences and being susceptible to semantic hallucination and popularity bias. To overcome these challenges, we propose RosePO, a framework to refine LLM-based recommendation through pairwise preference optimization with personalized smoothing. We illustrate the concept with three critical preference examples pertinent to LLM-based recommendation. Specifically, we design rejected sampling strategies tailored for each customized preference. To ensure robustness against uncertain labels present in automatically constructed preference data, we incorporate a personalized smoothing factor predicted by a user oracle into the optimization objective. Empirical evaluation on three real-world datasets demonstrates the effectiveness of our method, showcasing not only promising recommendation performance but also mitigation of semantic hallucination and popularity bias. We hope this work paves a way to build helpful and harmless LLM-based recommendation service in the future.

Empowering Compact LLMs with Fusion of Layer-wise Exits for Recommendation

Large language model-based recommender systems (LLM-RSs) have demonstrated remarkable capabilities, but are computationally unsustainable for many real-world applications. Compact LLMs offer a practical alternative, yet their reduced capacity often requires reasoning or knowledge distillation methods that increase latency or depend on larger models. Combined with autoregressive generation, these approaches face severe scalability bottlenecks. In contrast, discriminative LLM-RSs enable efficient full-corpus ranking through embedding similarity, but compact backbones remain limited in expressiveness and structural adaptivity. We propose the Fusion of Layer-wise Exits for Sequential Recommendation (FLEXRec), a discriminative framework that enhances compact LLMs while retaining scalable full-corpus ranking. FLEXRec inserts prediction heads (i.e., exits) at multiple transformer layers and adaptively fuses their score distributions. An adaptive continuous router (AC-Router) dynamically selects both the number and identity of exits for each user sequence, while a novel target-k hinge loss regulates routing sparsity. Experiments on three real-world datasets with Qwen 3 1.7B and Llama 3.2 3B show that FLEXRec achieves state-of-the-art accuracy among competing methods while remaining highly efficient. Code: https://github.com/xurong-liang/FLEXRec

PromptPack: Scaling LLM Annotation Agents for Online Recommendation

This work introduces PromptPack, a scalable, high-throughput LLM annotation agent that cuts LLM costs by 89% and accelerates throughput by 2.5x while fully preserving AUC.

Bridging the Semantic Gap: Leveraging LLMs for Hierarchical Interest Evolution in Sequential Recommendation

Preserving Item Semantics for Free: Rethinking Token Initialization in LLM-Based Generative Recommendation

Recent advances in generative recommendation (GR) leverage large language models (LLMs) as recommender backbones, enabling LLMs to directly generate recommendations conditioned on item-interaction histories. In these systems, items are often represented through semantic IDs (SIDs) added to the LLM vocabulary as special tokens. Ideally, SIDs imbue item token representations with semantic priors, thereby improving model generalization. However, standard vocabulary expansion typically initializes these tokens as random Gaussian vectors, discarding the SIDs'underlying continuous geometry and forcing the LLM to relearn token relationships from interaction data. To demonstrate the consequences of this design, we first show that training from this initialization tends to organize SID embeddings around item popularity rather than semantics. We further show that, despite partially reducing the reliance on popularity and improving cold item performance, the computationally expensive process of continual pretraining (CPT) fails to reliably recover the original semantic geometry. To address these findings, we propose a simple, parameter-free intervention that initializes SID token embeddings directly from their corresponding centroids in the semantic embedding space. Requiring only a few lines of code and no additional training or inference overhead, this drop-in approach improves pure-SFT Recall@5 by up to 16%, reaches peak performance with up to 40% fewer SFT steps, and improves cold-item Recall@5 by up to 60%. Moreover, on datasets that benefit from additional CPT, centroid initialization reaches comparable performance while requiring half as many CPT epochs. Together, our findings show that preserving SID geometry, beyond shared-prefix structure, provides a simple and effective semantic prior for LLM-based GR.

Pairwise Ranking Outperforms Single-Action RL for Offline Explanation Selection: A Practical Lesson

This work separates generation from selection: explanations are produced ahead of time as a frozen candidate pool (six prompt styles, two commodity LLMs), and a small CPU-resident selector picks one at request time, which needs no GPU and returns in under 100 ms.