RecRec: Latent Interests Recursive Reasoning for Sequential Recommendation
2026-07-14 • Information Retrieval
Information Retrieval
AI summaryⓘ
The authors study how recommendation systems predict what users might like next based on their past behavior. They note that previous methods try to combine reasoning and prediction into a single compact state, which limits how deeply the system can think before making recommendations. To fix this, the authors propose RecRec, which separates reasoning from prediction by first compressing user data into multiple distinct interests and then reasoning about these interests in a separate step. Their approach allows flexible reasoning depth without complicated training and performs better than prior methods on several real-world datasets.
Sequential recommender systemLatent reasoningReinforcement learningContext compressorLatent interestsRecursive reasonerDeep supervisionUser behavior modelingInference-time computationMulti-vector representation
Authors
Wenhao Deng, Junchen Fu, Hanwen Du, Alexandros Karatzoglou, Ioannis Arapakis, Hangjun Guo, Kaiwen Zheng, Yongxin Ni, Joemon M. Jose
Abstract
Sequential recommender systems rely on a single forward pass to encode user interaction histories and predict the next item. Increasing inference-time computation through latent reasoning, with the model proceeding step by step before the final prediction, has been recently explored in sequential recommendation with promising results. However, how to structure the reasoning process for sequential recommendation remains an open question. Existing approaches couple reasoning and prediction in a single $d$-dimensional state, limiting reasoning depth and often relying on multi-stage pipelines with reinforcement learning (RL). We propose RecRec (Recursive Reasoning for Recommendation), an RL-free framework that decouples reasoning from prediction, overcoming the fixed $d$-dimensional state bottleneck of prior methods. RecRec consists of a Context Compressor and a Recursive Reasoner, trained in two simple supervised stages. The Context Compressor distills the backbone's hidden states into a small set of latent interests, with an Interest Diversity Regularizer encouraging each interest to capture a distinct aspect of user behavior. The Recursive Reasoner then refines these interests by reasoning in a separate intermediate latent space. Deep supervision lets the reasoning depth be freely adjusted at inference without retraining. On four real-world datasets, RecRec outperforms state-of-the-art reasoning-enhanced methods, and on three of four datasets, gains extend past the training-time depth. Our findings point to a decoupled, multi-vector recipe that unleashes latent reasoning from the single-state bottleneck of prior methods, suggesting reasoning-state structure as a design axis to explore further in sequential recommendation.