arXiv subjects

Tao Di

Publications and source records attributed to Tao Di.

5 recordsLinked to original sources

arXiv2026

Learning Better Reasoning for Generative Recommendation with Semantic IDs

Generative recommendation reformulates item retrieval as sequence generation, allowing a unified model to directly generate the next item from a user's interaction history. Semantic IDs further make this paradigm effective and scalable by representing each item as discrete codes, enabling knowledge sharing among semantically related items. Recent studies introduce explicit reasoning before Semantic-ID generation, helping models summarize user interests and infer possible preference transitions. However, reasoning is not inherently beneficial: Inaccurate or uninformative reasoning may mislead subsequent item generation and ultimately degrade recommendation performance. This raises a central challenge: how can a recommender select and learn effective reasoning traces and progressively evolve toward better reasoning from its own generations? In this work, we propose Evo-Rec, a three-stage framework for learning better reasoning and further enhancing it through reinforcement learning. First, we align Semantic IDs with their textual and behavioral contexts, enabling the model to understand and generate item identifiers. Second, we sample multiple candidate reasoning traces and retain those that improve the prediction of the ground-truth item, providing a stronger reasoning initialization through supervised fine-tuning. Third, we further optimize the reasoning policy through reinforcement learning with catalog-constrained item generation and ranking-aware recommendation feedback. Experiments on three Amazon Review benchmarks show that Evo-Rec consistently outperforms discriminative, generative, and reasoning-enhanced recommenders across all evaluation metrics. These results demonstrate the effectiveness of our framework in learning better reasoning for SID-based generative recommendation.

cs.IR↗

arXiv2026

From Interests to Semantic IDs: Retrieval-Grounded Credit Assignment for Generative Recommendation

Semantic IDs (SIDs) encode each catalog item as a short token sequence, enabling generative recommenders to predict the next item autoregressively. Reasoning-enhanced variants, an increasingly common extension, first generate a textual trace and then decode a next-item SID by beam search. Such recommenders are commonly trained with group-relative policy optimization under an exact-match SID reward, which is sparse in large catalogs. Two failure modes follow. When all rollouts in a group miss the target, the group yields zero advantage and no learning signal. Rollouts sharing the same SID reward receive identical advantages, however much their traces differ. In both cases the reward reflects only the decoded SID, never the reasoning that produced it. This creates a credit-assignment gap. We address this gap with retrieval-grounded query attribution. Each trace is structured into a history summary, a set of interest hypotheses, and a final SID. A frozen retriever executes every hypothesis as a catalog query, so that each hypothesis becomes independently verifiable rather than judged only through the final SID. A rollout is rewarded when any of its queries retrieves the target within the \mbox{top-$K$}, and per-query hit indicators localize that reward to individual hypotheses. Credit is thus assigned at the span level: only hypotheses that individually hit receive positive retrieval advantage, while the retrieval channel never updates the final SID span. Rollouts that share a SID reward can therefore receive different updates. Across experiments on three Amazon Reviews datasets, this yields consistent improvements in SID recommendation. On Video Games, an oracle analysis further reveals the potential of interest-conditioned SID decoding: selecting the target-relevant query among generated interests improves both recall and ranking.

cs.IR↗

arXiv2026

Learning User Interests via Reasoning and Distillation for Cross-Domain News Recommendation

News recommendation plays a critical role in online news platforms by helping users discover relevant content. Cross-domain news recommendation further requires inferring user's underlying information needs from heterogeneous signals that often extend beyond direct news consumption. A key challenge lies in moving beyond surface-level behaviors to capture deeper, reusable user interests while maintaining scalability in large-scale production systems. In this paper, we present a reinforcement learning framework that trains large language models to generate high-quality lists of interest-driven news search queries from cross-domain user signals. We formulate query-list generation as a policy optimization problem and employ GRPO with multiple reward signals. We systematically study two compute dimensions: inference-time sampling and model capacity, and empirically observe consistent improvements with increased compute that exhibit scaling-like behavior. Finally, we perform on-policy distillation to transfer the learned policy from a large, compute-intensive teacher to a compact student model suitable for scalable deployment. Extensive offline experiments, ablation studies and large-scale online A/B tests in a production news recommendation system demonstrate consistent gains in both interest modeling quality and downstream recommendation performance.

cs.CL↗

arXiv2021

Training Large-Scale News Recommenders with Pretrained Language Models in the Loop

News recommendation calls for deep insights of news articles' underlying semantics. Therefore, pretrained language models (PLMs), like BERT and RoBERTa, may substantially contribute to the recommendation quality. However, it's extremely challenging to have news recommenders trained together with such big models: the learning of news recommenders requires intensive news encoding operations, whose cost is prohibitive if PLMs are used as the news encoder. In this paper, we propose a novel framework, {SpeedyFeed}, which efficiently trains PLMs-based news recommenders of superior quality. SpeedyFeed is highlighted for its light-weighted encoding pipeline, which gives rise to three major advantages. Firstly, it makes the intermedia results fully reusable for the training workflow, which removes most of the repetitive but redundant encoding operations. Secondly, it improves the data efficiency of the training workflow, where non-informative data can be eliminated from encoding. Thirdly, it further saves the cost by leveraging simplified news encoding and compact news representation. Extensive experiments show that SpeedyFeed leads to more than 100$\times$ acceleration of the training process, which enables big models to be trained efficiently and effectively over massive user data. The well-trained PLMs-based model from SpeedyFeed demonstrates highly competitive performance, where it outperforms the state-of-the-art news recommenders with significant margins. SpeedyFeed is also a model-agnostic framework, which is potentially applicable to a wide spectrum of content-based recommender systems; therefore, the whole framework is open-sourced to facilitate the progress in related areas.

cs.IR↗

arXiv2020

FedCTR: Federated Native Ad CTR Prediction with Multi-Platform User Behavior Data

Native ad is a popular type of online advertisement which has similar forms with the native content displayed on websites. Native ad CTR prediction is useful for improving user experience and platform revenue. However, it is challenging due to the lack of explicit user intent, and users' behaviors on the platform with native ads may not be sufficient to infer their interest in ads. Fortunately, user behaviors exist on many online platforms and they can provide complementary information for user interest mining. Thus, leveraging multi-platform user behaviors is useful for native ad CTR prediction. However, user behaviors are highly privacy-sensitive and the behavior data on different platforms cannot be directly aggregated due to user privacy concerns and data protection regulations like GDPR. Existing CTR prediction methods usually require centralized storage of user behavior data for user modeling and cannot be directly applied to the CTR prediction task with multi-platform user behaviors. In this paper, we propose a federated native ad CTR prediction method named FedCTR, which can learn user interest representations from their behaviors on multiple platforms in a privacy-preserving way. On each platform a local user model is used to learn user embeddings from the local user behaviors on that platform. The local user embeddings from different platforms are uploaded to a server for aggregation, and the aggregated user embeddings are sent to the ad platform for CTR prediction. Besides, we apply LDP and DP techniques to the local and aggregated user embeddings respectively for better privacy protection. Moreover, we propose a federated framework for model training with distributed models and user behaviors. Extensive experiments on real-world dataset show that FedCTR can effectively leverage multi-platform user behaviors for native ad CTR prediction in a privacy-preserving manner.

cs.IR↗