论文《The Oracle and The Prism》：解耦蒸馏框架 Prism 用于高效生成式推荐解释（已撤回）
The Oracle and The Prism: A Decoupled and Efficient Framework for Generative Recommendation Explanation
AI 导读
作者提出解耦框架 Prism，将可解释推荐拆分为排序与解释生成两个阶段，以 FLAN-T5-XXL 作为教师（Oracle）蒸馏知识给 140M 参数的 BART-Base 学生模型，实现 24x 加速和 10x 内存降低，并在忠实性与个性化人工评估中超过 11B 参数教师。该论文已由作者 Jiaheng Zhang 撤回，arXiv 页面注明论文存在若干问题。
正文 · 原文
This paper has been withdrawn by Jiaheng Zhang
No PDF available, click to view other formats
Abstract:The integration of Large Language Models (LLMs) into explainable recommendation systems often leads to a performance-efficiency trade-off in end-to-end architectures, where joint optimization of ranking and explanation can result in suboptimal compromises. To resolve this, we propose Prism, a novel decoupled framework that rigorously separates the recommendation process into a dedicated ranking stage and an explanation generation stage. This decomposition ensures that each component is optimized for its specific objective, eliminating inherent conflicts in coupled models.
Inspired by knowledge distillation, Prism leverages a powerful, instruction-following teacher LLM (FLAN-T5-XXL) as an Oracle to produce high-fidelity explanatory knowledge. A compact, fine-tuned student model (BART-Base), the Prism, then specializes in synthesizing this knowledge into personalized explanations. Our extensive experiments on benchmark datasets reveal a key finding: the distillation process not only transfers knowledge but also acts as a noise filter. Our 140M-parameter Prism model significantly outperforms its 11B-parameter teacher in human evaluations of faithfulness and personalization, demonstrating an emergent ability to correct hallucinations present in the teacher's outputs. While achieving a 24x speedup and a 10x reduction in memory consumption, our analysis validates that decoupling, coupled with targeted distillation, provides an efficient and effective pathway to high-quality, and perhaps more importantly, trustworthy explainable recommendation.
| Comments: | We have identified several issues in the paper |
| Subjects: | Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG) |
| Cite as: | arXiv:2511.16543 [cs.IR] |
| (or arXiv:2511.16543v3 [cs.IR] for this version) | |
| https://doi.org/10.48550/arXiv.2511.16543 arXiv-issued DOI via DataCite |
Submission history
From: Jiaheng Zhang [view email]
[v1]
Thu, 20 Nov 2025 16:59:16 UTC (2,231 KB)
[v2]
Sat, 6 Dec 2025 08:27:39 UTC (2,232 KB)
[v3]
Wed, 11 Nov 2026 13:41:40 UTC (1 KB) (withdrawn)
来源：arXiv cs.AI · arxiv.org