arXiv subjects

Hailong Shi

Publications and source records attributed to Hailong Shi.

10 recordsLinked to original sources

arXiv2026

ChainRec: An Agentic Recommender Learning to Route Tool Chains for Diverse and Evolving Interests

Large language models (LLMs) are increasingly integrated into recommender systems, motivating recent interest in agentic and reasoning-based recommendation. However, most existing approaches still rely on fixed workflows, applying the same reasoning procedure across diverse recommendation scenarios. In practice, user contexts vary substantially-for example, in cold-start settings or during interest shifts, so an agent should adaptively decide what evidence to gather next rather than following a scripted process. To address this, we propose ChainRec, an agentic recommender that uses a planner to dynamically select reasoning tools. ChainRec builds a standardized Tool Agent Library from expert trajectories. It then trains a planner using supervised fine-tuning and preference optimization to dynamically select tools, decide their order, and determine when to stop. Experiments on AgentRecBench across Amazon, Yelp, and Goodreads show that ChainRec consistently improves Avg HR@{1,3,5} over strong baselines, with especially notable gains in cold-start and evolving-interest scenarios. Ablation studies further validate the importance of tool standardization and preference-optimized planning.

cs.IR↗

arXiv2025

SCoTER: Structured Chain-of-Thought Transfer for Enhanced Recommendation

Harnessing the reasoning power of Large Language Models (LLMs) for recommender systems is hindered by two fundamental challenges. First, current approaches lack a mechanism for automated, data-driven discovery of effective reasoning patterns, relying instead on brittle manual templates or unstable zero-shot prompting. Second, they employ structure-collapsing integration: direct prompting incurs prohibitive online inference costs, while feature extraction collapses reasoning chains into single vectors, discarding stepwise logic. To address these challenges, we propose SCoTER (Structured Chain-of-Thought Transfer for Enhanced Recommendation), a unified framework that treats pattern discovery and structure-aware transfer as a jointly optimized problem. Specifically, SCoTER operationalizes this through two synergistic components: a Generate-Validate-Mine (GVM) pipeline for automated pattern discovery and a structure-preserving integration architecture that transfers stepwise logic to efficient models. Empirically, experiments on four benchmarks demonstrate consistent improvements across diverse backbones. Moreover, in production deployment on the Tencent Advertising Platform, SCoTER achieved a 2.14\% lift in Gross Merchandise Value (GMV) while eliminating online LLM inference costs. Overall, SCoTER presents a practical and unified framework for integrating structured LLM reasoning into recommender systems, validated by consistent improvements in both offline benchmarks and online production environments.

cs.IR↗

arXiv2025

Efficient Parallel Training Methods for Spiking Neural Networks with Constant Time Complexity

Spiking Neural Networks (SNNs) often suffer from high time complexity $O(T)$ due to the sequential processing of $T$ spikes, making training computationally expensive. In this paper, we propose a novel Fixed-point Parallel Training (FPT) method to accelerate SNN training without modifying the network architecture or introducing additional assumptions. FPT reduces the time complexity to $O(K)$, where $K$ is a small constant (usually $K=3$), by using a fixed-point iteration form of Leaky Integrate-and-Fire (LIF) neurons for all $T$ timesteps. We provide a theoretical convergence analysis of FPT and demonstrate that existing parallel spiking neurons can be viewed as special cases of our proposed method. Experimental results show that FPT effectively simulates the dynamics of original LIF neurons, significantly reducing computational time without sacrificing accuracy. This makes FPT a scalable and efficient solution for real-world applications, particularly for long-term tasks. Our code will be released at \href{https://github.com/WanjinVon/FPT}{\texttt{https://github.com/WanjinVon/FPT}}.

cs.NE↗

arXiv2023

An Adaptive Framework of Geographical Group-Specific Network on O2O Recommendation

Online to offline recommendation strongly correlates with the user and service's spatiotemporal information, therefore calling for a higher degree of model personalization. The traditional methodology is based on a uniform model structure trained by collected centralized data, which is unlikely to capture all user patterns over different geographical areas or time periods. To tackle this challenge, we propose a geographical group-specific modeling method called GeoGrouse, which simultaneously studies the common knowledge as well as group-specific knowledge of user preferences. An automatic grouping paradigm is employed and verified based on users' geographical grouping indicators. Offline and online experiments are conducted to verify the effectiveness of our approach, and substantial business improvement is achieved.

cs.IR↗

arXiv2015

The Volume-Correlation Subspace Detector

Detecting the presence of subspace signals with unknown clutter (or interference) is a widely known difficult problem encountered in various signal processing applications. Traditional methods fails to solve this problem because they require knowledge of clutter subspace, which has to be learned or estimated beforehand. In this paper, we propose a novel detector, named volume-correlation subspace detector, that can detect signal from clutter without any knowledge of clutter subspace. This detector effectively makes use of the hidden geometrical connection between the known target signal subspace to be detected and the subspace constructed from sampled data to ascertain the existence of target signal. It is derived based upon a mathematical tool, which basically calculates volume of parallelotope in high-dimensional linear space. Theoretical analysis show that while the proposed detector is detecting the known target signal, the unknown clutter signal can be explored and eliminated simultaneously. This advantage is called "detecting while learning", and implies perfect performance of this detector in the clutter environment. Numerical simulation validated our conclusion.

cs.IT↗

arXiv2015

A TDOA technique with Super-Resolution based on the Volume Cross-Correlation Function

Time Difference of Arrival (TDOA) is widely used in wireless localization systems. Among the enormous approaches of TDOA, high resolution TDOA algorithms have drawn much attention for its ability to resolve closely spaced signal delays in multipath environment. However, the state-of-art high resolution TDOA algorithms still have performance weakness on resolving time delays in a wireless channel with dense multipath effect, as well as difficulties in implementation for their high computation complexity. In this paper, we propose a novel TDOA algorithm with super resolution based on a multi-dimensional cross-correlation function: the Volume Cross-Correlation Function (VCC). The proposed TDOA algorithm has excellent time resolution capability in multipath environment, and it also has a much lower computational complexity. Because our algorithm does not require priori knowledge about the waveform or power spectrum of transmitted signals, it has great potential of usage in various passive wireless localization systems. Numerical simulations is also provided to demonstrate the validity of our conclusion.

cs.IT↗

arXiv2015

A Volume Correlation Subspace Detector for signals buried in unknown clutter

Detecting the presence of target subspace signals with unknown clutters is a well-known hard problem encountered in various signal processing applications. Traditional methods fails to solve this problem because prior knowledge of clutter subspace is required, which can not be obtained when target and clutter are intimately mixed. In this paper, we propose a novel subspace detector that can detect target signal buried in clutter without knowledge of clutter subspace. This detector makes use of the geometrical relation between target and clutter subspaces and is derived based upon the calculation of volume of high dimensional geometrical objects. Moreover, the proposed detector can accomplish the detection simultaneously with the learning processes of clutter, a property called "detecting while learning". The performance of detector was showed by theoretical analysis and numerical simulation.

cs.IT↗

arXiv2014

Stable Embedding of Grassmann Manifold via Gaussian Random matrices

In this paper, we explore a volume-based stable embedding of multi-dimensional signals based on Grassmann manifold, via Gaussian random measurement matrices. The Grassmann manifold is a topological space in which each point is a linear vector subspace, and is widely regarded as an ideal model for multi-dimensional signals. In this paper, we formulate the linear subspace spanned by multi-dimensional signal vectors as points on the Grassmann manifold, and use the volume and the product of sines of principal angles (also known as the product of principal sines) as the generalized norm and distance measure for the space of Grassmann manifold. We prove a volume-preserving embedding property for points on the Grassmann manifold via Gaussian random measurement matrices, i.e., the volumes of all parallelotopes from a finite set in Grassmann manifold are preserved upon compression. This volume-preserving embedding property is a multi-dimensional generalization of the conventional stable embedding properties, which only concern the approximate preservation of lengths of vectors in certain unions of subspaces. Additionally, we use the volume-preserving embedding property to explore the stable embedding effect on a generalized distance measure of Grassmann manifold induced from volume. It is proved that the generalized distance measure, i.e., the product of principal sines between different points on the Grassmann manifold, is well preserved in the compressed domain via Gaussian random measurement matrices.Numerical simulations are also provided for validation.

cs.IT↗

arXiv2013

Quasi-Equiangular Frame (QEF) : A New Flexible Configuration of Frame

Frame theory is a powerful tool in the domain of signal processing and communication. Among its numerous configurations, the ones which have drawn much attention recently are Equiangular Tight Frame (ETF) and Grassmannian Frame. These frames both have some kind of optimality in coherence, thus bring robustness or optimal performance in applications such as digital fingerprint, erasure channels, and Compressive Sensing. However, too strict constraint on existence and construction of ETF and Grassmannian Frame became the main obstacle for widespread use. In this paper, we propose a new configuration of frame: Quasi-Equiangular Frame, as a compromise but more convenient and flexible approximation of ETF and Grassmannian Frame. We will give formal definition of Quasi-Equiangular Frame and analyze its relationship with ETF and Grassmannian frame. Furthermore, for popularity of ETF and Grassmannian frame in Compressive Sensing, we utilize the technique of random matrices to obtain asymptotical concentration estimation of the Restricted Isometry Constant (RIC) of Quasi-Equiangular Frame with respect to its key parameter.

cs.IT↗

arXiv2013

Tight is better: Performance Improvement of the Compressive Classifier Using Equi-Norm Tight Frames

Detecting or classifying already known sparse signals contaminated by Gaussian noise from compressive measurements is different from reconstructing sparse signals, as its objective is to minimize the error probability which describes performance of the detectors or classifiers. This paper is concerned about the performance improvement of a commonly used Compressive Classifier. We prove that when the arbitrary sensing matrices used to get the Compressive Measurements are transformed into Equi-Norm Tight Frames, i.e. the matrices that are row-orthogonal, The Compressive Classifier achieves better performance. Although there are other proofs that among all Equi-Norm Tight Frames the Equiangular tight Frames (ETFs) bring best worst-case performance, the existence and construction of ETFs on some dimensions is still an open problem. As the construction of Equi-Norm Tight Frames from any arbitrary matrices is very easy and practical compared with ETF matrices, the result of this paper can also provide a practical method to design an improved sensing matrix for Compressive Classification. We can conclude that: Tight is Better!

cs.IT↗