Search papers, labs, and topics across Lattice.
The paper introduces FeDecider, a federated cross-domain recommendation framework that leverages LLMs while addressing challenges of overfitting and heterogeneous cross-domain similarity measurement. FeDecider disentangles low-rank updates and shares only directional components to mitigate scale-specific noise during federated aggregation. Furthermore, it learns personalized weights for each client to enable data-aware integration of updates from other domains, improving recommendation performance across diverse datasets.
Federated learning with LLMs for recommendation can be improved by sharing directional update components and learning personalized integration weights, outperforming naive aggregation.
Federated cross-domain recommendation (Federated CDR) aims to collaboratively learn personalized recommendation models across heterogeneous domains while preserving data privacy. Recently, large language model (LLM)-based recommendation models have demonstrated impressive performance by leveraging LLMs' strong reasoning capabilities and broad knowledge. However, adopting LLM-based recommendation models in Federated CDR scenarios introduces new challenges. First, there exists a risk of overfitting with domain-specific local adapters. The magnitudes of locally optimized parameter updates often vary across domains, causing biased aggregation and overfitting toward domain-specific distributions. Second, unlike traditional recommendation models (e.g., collaborative filtering, bipartite graph-based methods) that learn explicit and comparable user/item representations, LLMs encode knowledge implicitly through autoregressive text generation training. This poses additional challenges for effectively measuring the cross-domain similarities under heterogeneity. To address these challenges, we propose an LLM-based framework for federated cross-domain recommendation, FeDecider. Specifically, FeDecider tackles the challenge of scale-specific noise by disentangling each client's low-rank updates and sharing only their directional components. To handle the need for flexible and effective integration, each client further learns personalized weights that achieve the data-aware integration of updates from other domains. Extensive experiments across diverse datasets validate the effectiveness of our proposed FeDecider.
