arXiv subjects

Charlie Xu

Publications and source records attributed to Charlie Xu.

3 recordsLinked to original sources

arXiv2025

Continual Low-Rank Adapters for LLM-based Generative Recommender Systems

While large language models (LLMs) achieve strong performance in recommendation, they face challenges in continual learning as users, items, and user preferences evolve over time. Existing LoRA-based continual methods primarily focus on preserving performance on previous tasks, but this overlooks the unique nature of recommendation: the goal is not to predict past preferences, and outdated preferences can even harm performance when current interests shift significantly. To address this, we propose PESO (Proximally rEgularized Single evolving lOra, a continual adaptation method for LoRA in recommendation. PESO introduces a proximal regularizer that anchors the current adapter to its most recent frozen state, enabling the model to flexibly balance adaptation and preservation, and to better capture recent user behaviors. Theoretically, we show that this proximal design provides data-aware, direction-wise guidance in the LoRA subspace. Empirically, PESO consistently outperforms existing LoRA-based continual learning methods.

cs.LG↗

arXiv2023

Ensuring User-side Fairness in Dynamic Recommender Systems

User-side group fairness is crucial for modern recommender systems, aiming to alleviate performance disparities among user groups defined by sensitive attributes like gender, race, or age. In the ever-evolving landscape of user-item interactions, continual adaptation to newly collected data is crucial for recommender systems to stay aligned with the latest user preferences. However, we observe that such continual adaptation often exacerbates performance disparities. This necessitates a thorough investigation into user-side fairness in dynamic recommender systems, an area that has been unexplored in the literature. This problem is challenging due to distribution shifts, frequent model updates, and non-differentiability of ranking metrics. To our knowledge, this paper presents the first principled study on ensuring user-side fairness in dynamic recommender systems. We start with theoretical analyses on fine-tuning v.s. retraining, showing that the best practice is incremental fine-tuning with restart. Guided by our theoretical analyses, we propose FAir Dynamic rEcommender (FADE), an end-to-end fine-tuning framework to dynamically ensure user-side fairness over time. To overcome the non-differentiability of recommendation metrics in the fairness loss, we further introduce Differentiable Hit (DH) as an improvement over the recent NeuralNDCG method, not only alleviating its gradient vanishing issue but also achieving higher efficiency. Besides that, we also address the instability issue of the fairness loss by leveraging the competing nature between the recommendation loss and the fairness loss. Through extensive experiments on real-world datasets, we demonstrate that FADE effectively and efficiently reduces performance disparities with little sacrifice in the overall recommendation performance.

cs.IR↗

arXiv2022

Installation, Commissioning and Performance of Phase Reference Line for LCLS-II

Any cavity controller for a distributed system needs a Phase Reference Line (PRL) signal from which to define phases of a cavity field measurement. The LCLS-II PRL system at SLAC provides bidirectional (forward and reverse) phase references at 1300 MHz to each rack of the LLRF system. The PRL controller embedded with the Master Oscillator (MO) locks the average phase of the two directions to the MO itself. Phase-averaging tracking loop is applied in firmware which supports the feature of cancelling the phase drift caused by changes in PRL cable length. FPGA logic moves the phase of digital LO to get zero average phase of the two PRL signals. This same LO is used for processing cavity pickup signals, thus establishing a stable reference phase for critical cavity RF measurements. At low frequencies, open-loop PRL noise relative to the LO distribution includes a strong environment and 1/f components, but the closed-loop noise approaches the noise floor of the DSP. The implication is that the close-in phase noise of the cavities will be dominated by the chassis DAQ noise. The final out-of-loop phase noise relevant to machine operations is that of the cavity field relative to the beam.

physics.acc-ph↗