aiqbench analysis
This research addresses critical bottlenecks in merging Large Language Models for cross-domain recommendation by introducing SharpRec, a framework that combines sharpness-aware geometric alignment with salience recovery to prevent parameter misalignment and knowledge saturation. By stabilizing the fusion process while preserving distinctive domain features, the method effectively lifts performance upper bounds without relying on overlapping user data. The approach represents a significant advancement in scalable multi-model integration for sequential recommendation tasks within machine learning systems.
Sharpness-aware Model Merging with Salience Recovery for LLM-based Cross-Domain Sequential Recommendation
arXiv:2607.25366v1 Announce Type: cross Abstract: LLM-based Cross-Domain Sequential Recommendation (CDSR) leverages LLMs to enhance target performance via deep semantic reasoning, alleviating the dependency on overlapping users. Among LLM-based paradigms, model merging is particularly promising for multi-domain scenarios due to its superior scalability and flexibility in integrating diverse knowledge sources. However, our empirical investigations reveal two critical bottlenecks: (1) cross-domain knowledge conflict; and (2) performance saturation in multi-domain fusion. Our analysis attributes these phenomena to parameter-level misalignment and statistical homogenization during the merging process. To address these bottlenecks, we propose SharpRec, Sharpness-aware Model Merging with Salience Recovery for LLM-based CDSR, a framework designed to lift the performance upper bound of merged models. SharpRec incorporates two synergistic modules: Sharpness-aware Geometric Alignment to establish a stable geometric foundation for interference-free fusion; and Preference Salience Activation to effectively recover the distinctive features essential for bolstering target domain performance. Extensive experiments in both dual-domain and multi-domain scenarios demonstrate that SharpRec consistently outperforms state-of-the-art baselines.
Abstract:LLM-based Cross-Domain Sequential Recommendation (CDSR) leverages LLMs to enhance target performance via deep semantic reasoning, alleviating the dependency on overlapping users. Among LLM-based paradigms, model merging is particularly promising for multi-domain scenarios due to its superior scalability and flexibility in integrating diverse knowledge sources. However, our empirical investigations reveal two critical bottlenecks: (1) cross-domain knowledge conflict; and (2) performance saturation in multi-domain fusion. Our analysis attributes these phenomena to parameter-level misalignment and statistical homogenization during the merging process. To address these bottlenecks, we propose SharpRec, Sharpness-aware Model Merging with Salience Recovery for LLM-based CDSR, a framework designed to lift the performance upper bound of merged models. SharpRec incorporates two synergistic modules: Sharpness-aware Geometric Alignment to establish a stable geometric foundation for interference-free fusion; and Preference Salience Activation to effectively recover the distinctive features essential for bolstering target domain performance. Extensive experiments in both dual-domain and multi-domain scenarios demonstrate that SharpRec consistently outperforms state-of-the-art baselines.
Published in Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD '26). 12 pages, 6 figures, 4 tables. Code available at this https URL
Information Retrieval (cs.IR); Machine Learning (cs.LG)
arXiv:2607.25366 [cs.IR]
(or arXiv:2607.25366v1 [cs.IR] for this version)
https://doi.org/10.48550/arXiv.2607.25366 arXiv-issued DOI via DataCite (pending registration)
Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2 (KDD '26), August 9-13, 2026, Jeju Island, Republic of Korea, ACM, 2026
https://doi.org/10.1145/3770855.3817945 DOI(s) linking to related resources
Full text under the publisher's open licence · 236 words
Related coverage
J++ Lens: Jacobian Filtering Enables More Faithful Workspace Lenses
LessWrong · Oct 8, 2026
When Privacy Hurts Mergeability: Geometry-Aware Model Merging under Differential Privacy
arXiv · Aug 28, 2026
Why do models task game?
AI Alignment Forum · Aug 6, 2026
Stock Market Today Highlights, Sept 2: Sensex ends 373 pts lower at 76, 570, Nifty below 24, 000; Asian Paints, HDFC Bank fall most
thehindubusinessline.com · Sep 2, 2026