This work proposes the first model merging framework for reasoning compression in recommender systems, and proposes selective injection of the concise behaviour of the fast-thinking model into the slow-thinking model and reducing reasoning verbosity without compromising recommendation quality.
Abstract
Large language model-based recommender systems are increasingly adopting slow-thinking models that generate step-by-step reasoning before making predictions, often achieving higher accuracy than fast-thinking models that predict directly. However, their reasoning traces are often unnecessarily verbose, increasing inference costs without commensurate accuracy gains. Existing training-based approaches to reasoning compression often incur substantial adaptation costs, while inference-time methods are brittle and difficult to scale. These limitations motivate model merging as a promising training-free direction for transferring specialised behaviours between models in a shared parameter space. In particular, merging a slow-thinking model with a fast-thinking counterpart provides a natural mechanism for balancing recommendation accuracy and reasoning conciseness. To this end, we propose, to our knowledge, the first model merging framework for reasoning compression in recommender systems. Unlike conventional merging methods that apply uniform merge coefficients across model components, our method performs fine-grained merging at the level of individual attention heads, capturing heterogeneous patterns in recommendation reasoning. Each attention head is assigned a distinct merge coefficient according to its contribution to critical reasoning evidence and its sensitivity to parameter change, enabling selective injection of the concise behaviour of the fast-thinking model into the slow-thinking model and reducing reasoning verbosity without compromising recommendation quality. Experiments on three benchmark datasets show that our method reduces reasoning length by up to 24.3% while outperforming competitive model merging baselines in maintaining recommendation accuracy. The code is available at https://github.com/linhledieu/REAM.
Large Language Models (LLMs) have recently emerged as powerful backbones for recommendation. To better elicit their capabilities, reasoning has been widely incorporated to help LLMs interpret rich textual signals and improve recommendation accuracy. However, explicitly generating intermediate reasoning traces often incurs substantial computational costs, which limits practical deployment in real-world recommender systems. To address this challenge, we propose SelfDR, a Self-Distillation from Reasoning framework for LLM-based Recommendation. SelfDR distills an LLM's own reasoning-enhanced predictions to produce recommendations directly, improving recommendation effectiveness while maintaining inference efficiency. All components in the framework are built on the same base LLM, without relying on any external models. Specifically, the teacher recommender is constructed by training a reasoner with downstream performance as the reward, enabling it to generate targeted rationales that are later incorporated into the teacher's input. A student recommender for direct recommendation, with the same underlying model, then learns from the teacher through self-distillation with a dynamic weighting strategy. Extensive experiments on three public datasets validate the effectiveness, rationality, and efficiency of SelfDR. Codes are available at https://github.com/JiangDeccc/SelfDistillation.
Chumeng Jiang, Jiayin Wang, Xin-Jie Lin et al.· 0 citations
Large Language Models (LLMs) have demonstrated strong potential in recommendations due to their powerful reasoning capabilities. However, existing methods typically rely on a single reasoning path to drive the entire Top-K recommendations. This paradigm is prone to reasoning path collapse, where limiting exploration of potentially superior and diverse reasoning paths within the LLMs space. As a result, both the accuracy and diversity of the recommendation outcomes are constrained. To address this issue, we propose a novel model, Divergent Reasoning for LLM-based Recommendation, named DivReason. Inspired by the structure of intellect theory, which emphasizes a two-stage cognitive process of divergent thinking followed by convergent thinking, DivReason is designed with two core components: the Divergent Reasoning Path Generation Module and the Reasoning Path Aggregation Module. In the first module, DivReason introduces a training-free form of controlled uncertainty to promote diverse reasoning, leveraging Monte Carlo Dropout and Directional Perturbation to expand exploration in the latent reasoning space. In the Reasoning Path Aggregation Module, we adaptively select a subset of high-quality reasoning paths from the entire path pool and aggregate them into a unified reasoning representation. Meanwhile, we further adopt an alternating reinforcement learning strategy to optimize the model, explicitly balancing accuracy and diversity during training. Extensive experimental results show that DivReason effectively mitigates the issue of reasoning path collapse, while improving both the accuracy and diversity of LLM-based recommendations.
Guojia An, Jie Zou, Yuhan Yang et al.· Annual International ACM SIG...· 1 citation
Traditional sequential recommender systems have primarily relied on implicit pattern recognition in user interaction sequences, achieving strong performance but functioning as “black boxes” that lack transparent reasoning. This paper introduces LLM-RS, a novel framework that leverages Large Language Models to enable explicit reasoning chains in sequential recommendation. Our approach transforms the recommendation task from mere pattern matching to interpretable reasoning by developing a multi-stage architecture that: (1) extracts structured preference profiles from user interaction sequences, (2) generates explicit reasoning chains analyzing candidate items against inferred preferences, and (3) produces persuasive explanations alongside recommendations. We propose three model variants—fine-tuned reasoning, retrieval-augmented generation, and hybrid ensemble—that integrate LLM-based reasoning with traditional collaborative filtering. The framework addresses key challenges in modern recommender systems by providing transparent, persuasive rationales while maintaining competitive performance, marking a significant step toward more interpretable and trustworthy recommendation systems. Comprehensive evaluations across the Amazon Reviews, MovieLens, MIND, and KuaiSAR datasets demonstrate that LLM-RS not only matches state-of-the-art methods in accuracy but also significantly enhances explanation quality, user trust, and recommendation diversity. Our findings reveal that reasoning-enabled recommendations increase user adherence in online experiments and improve long-term engagement metrics.
A. Gadallah, H. Hefny, Mohammed E. Almandouh et al.· Mathematics· 0 citations
WhisperRec compresses teacher-generated CoT into learnable latent reasoning tokens, enabling a Latent-Reason-then-Answer paradigm that performs reasoning in latent space without producing verbose rationales, and achieves over 10x higher online inference throughput.
Hao Jiang, Pei Du, Pengfei Yao et al.· arXiv.org· 0 citations
To further optimize the reasoning trajectory, HiLaR combines final recommendation feedback with layer-aware process rewards derived from the marginal target-likelihood gain of each state, and generally outperforms strong sequential, generative, and LLM-based recommendation baselines.
Peiyu Hu, Siying Gu, Weihai Lu et al.· arXiv.org· 1 citation
I Reasoning via Multi-Teacher Distillation is proposed, a novel framework that 'compiles' the reasoning abilities of large teacher LLMs into a lightweight student Small Language Model (SLM), which significantly outperforms state-of-the-art baselines in both recommendation accuracy and inference efficiency.