Search papers, labs, and topics across Lattice.
This paper introduces a novel model merging framework for enhancing reasoning efficiency in large language model-based recommender systems by integrating slow-thinking and fast-thinking models. The approach addresses the verbosity of reasoning traces that inflate inference costs without significant accuracy improvements, achieving a reduction in reasoning length by up to 24.3% while preserving recommendation quality. By employing fine-grained merging at the level of individual attention heads, the method allows for selective incorporation of concise reasoning behaviors, outperforming existing baselines in both efficiency and accuracy.
Merging slow and fast-thinking models can cut reasoning verbosity by over 24% without sacrificing accuracy, revolutionizing LLM-based recommendations.
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.
