arXiv subjects

Shuai Lin

Publications and source records attributed to Shuai Lin.

11 recordsLinked to original sources

arXiv2026

Dual-Rerank: Fusing Causality and Utility for Industrial Generative Reranking

Kuaishou serves over 400 million daily active users, processing hundreds of millions of search queries daily against a repository of tens of billions of short videos. As the final decision layer, the reranking stage determines user experience by optimizing whole-page utility. While traditional score-and-sort methods fail to capture combinatorial dependencies, Generative Reranking offers a superior paradigm by directly modeling the permutation probability. However, deploying Generative Reranking in such a high-stakes environment faces a fundamental dual dilemma: 1) the structural trade-off where Autoregressive (AR) models offer superior Sequential modeling but suffer from prohibitive latency, versus Non-Autoregressive (NAR) models that enable efficiency but lack dependency capturing; 2) the optimization gap where Supervised Learning faces challenges in directly optimizing whole-page utility, while Reinforcement Learning (RL) struggles with instability in high-throughput data streams. To resolve this, we propose Dual-Rerank, a unified framework designed for industrial reranking that bridges the structural gap via Sequential Knowledge Distillation and addresses the optimization gap using List-wise Decoupled Reranking Optimization (LDRO) for stable online RL. Extensive A/B testing on production traffic demonstrates that Dual-Rerank achieves State-of-the-Art performance, significantly improving User satisfaction and Watch Time while drastically reducing inference latency compared to AR baselines.

cs.IR↗

arXiv2025

GReF: A Unified Generative Framework for Efficient Reranking via Ordered Multi-token Prediction

In a multi-stage recommendation system, reranking plays a crucial role in modeling intra-list correlations among items. A key challenge lies in exploring optimal sequences within the combinatorial space of permutations. Recent research follows a two-stage (generator-evaluator) paradigm, where a generator produces multiple feasible sequences, and an evaluator selects the best one. In practice, the generator is typically implemented as an autoregressive model. However, these two-stage methods face two main challenges. First, the separation of the generator and evaluator hinders end-to-end training. Second, autoregressive generators suffer from inference efficiency. In this work, we propose a Unified Generative Efficient Reranking Framework (GReF) to address the two primary challenges. Specifically, we introduce Gen-Reranker, an autoregressive generator featuring a bidirectional encoder and a dynamic autoregressive decoder to generate causal reranking sequences. Subsequently, we pre-train Gen-Reranker on the item exposure order for high-quality parameter initialization. To eliminate the need for the evaluator while integrating sequence-level evaluation during training for end-to-end optimization, we propose post-training the model through Rerank-DPO. Moreover, for efficient autoregressive inference, we introduce ordered multi-token prediction (OMTP), which trains Gen-Reranker to simultaneously generate multiple future items while preserving their order, ensuring practical deployment in real-time recommender systems. Extensive offline experiments demonstrate that GReF outperforms state-of-the-art reranking methods while achieving latency that is nearly comparable to non-autoregressive models. Additionally, GReF has also been deployed in a real-world video app Kuaishou with over 300 million daily active users, significantly improving online recommendation quality.

cs.IR↗

arXiv2021

Prototypical Graph Contrastive Learning

Graph-level representations are critical in various real-world applications, such as predicting the properties of molecules. But in practice, precise graph annotations are generally very expensive and time-consuming. To address this issue, graph contrastive learning constructs instance discrimination task which pulls together positive pairs (augmentation pairs of the same graph) and pushes away negative pairs (augmentation pairs of different graphs) for unsupervised representation learning. However, since for a query, its negatives are uniformly sampled from all graphs, existing methods suffer from the critical sampling bias issue, i.e., the negatives likely having the same semantic structure with the query, leading to performance degradation. To mitigate this sampling bias issue, in this paper, we propose a Prototypical Graph Contrastive Learning (PGCL) approach. Specifically, PGCL models the underlying semantic structure of the graph data via clustering semantically similar graphs into the same group, and simultaneously encourages the clustering consistency for different augmentations of the same graph. Then given a query, it performs negative sampling via drawing the graphs from those clusters that differ from the cluster of query, which ensures the semantic difference between query and its negative samples. Moreover, for a query, PGCL further reweights its negative samples based on the distance between their prototypes (cluster centroids) and the query prototype such that those negatives having moderate prototype distance enjoy relatively large weights. This reweighting strategy is proved to be more effective than uniform sampling. Experimental results on various graph benchmarks testify the advantages of our PGCL over state-of-the-art methods. Code is publicly available at https://github.com/ha-lins/PGCL.

cs.LG↗

arXiv2020

Graph-Evolving Meta-Learning for Low-Resource Medical Dialogue Generation

Human doctors with well-structured medical knowledge can diagnose a disease merely via a few conversations with patients about symptoms. In contrast, existing knowledge-grounded dialogue systems often require a large number of dialogue instances to learn as they fail to capture the correlations between different diseases and neglect the diagnostic experience shared among them. To address this issue, we propose a more natural and practical paradigm, i.e., low-resource medical dialogue generation, which can transfer the diagnostic experience from source diseases to target ones with a handful of data for adaptation. It is capitalized on a commonsense knowledge graph to characterize the prior disease-symptom relations. Besides, we develop a Graph-Evolving Meta-Learning (GEML) framework that learns to evolve the commonsense graph for reasoning disease-symptom correlations in a new disease, which effectively alleviates the needs of a large number of dialogues. More importantly, by dynamically evolving disease-symptom graphs, GEML also well addresses the real-world challenges that the disease-symptom correlations of each disease may vary or evolve along with more diagnostic cases. Extensive experiment results on the CMDD dataset and our newly-collected Chunyu dataset testify the superiority of our approach over state-of-the-art approaches. Besides, our GEML can generate an enriched dialogue-sensitive knowledge graph in an online manner, which could benefit other tasks grounded on knowledge graph.

cs.CL↗

arXiv2020

Iterative Graph Self-Distillation

Recently, there has been increasing interest in the challenge of how to discriminatively vectorize graphs. To address this, we propose a method called Iterative Graph Self-Distillation (IGSD) which learns graph-level representation in an unsupervised manner through instance discrimination using a self-supervised contrastive learning approach. IGSD involves a teacher-student distillation process that uses graph diffusion augmentations and constructs the teacher model using an exponential moving average of the student model. The intuition behind IGSD is to predict the teacher network representation of the graph pairs under different augmented views. As a natural extension, we also apply IGSD to semi-supervised scenarios by jointly regularizing the network with both supervised and self-supervised contrastive loss. Finally, we show that finetuning the IGSD-trained models with self-training can further improve the graph representation power. Empirically, we achieve significant and consistent performance gain on various graph datasets in both unsupervised and semi-supervised settings, which well validates the superiority of IGSD.

cs.LG↗

arXiv2019

Clark-Ocone Formula for Generalized Functionals of Discrete-Time Normal Noises

The Clark-Ocone formula in the theory of discrete-time chaotic calculus holds only for square integrable functionals of discrete-time normal noises. In this paper, we aim at extending this formula to generalized functionals of discrete-time normal noises. Let $Z$ be a discrete-time normal noise that has the chaotic representation property. We first prove a result concerning the regularity of generalized functionals of $Z$. Then, we use the Fock transform to define some fundamental operators on generalized functionals of $Z$, and apply the above mentioned regularity result to prove the continuity of these operators. Finally, we establish the Clark-Ocone formula for generalized functionals of $Z$, and show its application results, which include the covariant identity result and the variant upper bound result for generalized functionals of $Z$.

math.PR↗

arXiv2019

Data-to-Text Generation with Style Imitation

Recent neural approaches to data-to-text generation have mostly focused on improving content fidelity while lacking explicit control over writing styles (e.g., word choices, sentence structures). More traditional systems use templates to determine the realization of text. Yet manual or automatic construction of high-quality templates is difficult, and a template acting as hard constraints could harm content fidelity when it does not match the record perfectly. We study a new way of stylistic control by using existing sentences as soft templates. That is, the model learns to imitate the writing style of any given exemplar sentence, with automatic adaptions to faithfully describe the content record. The problem is challenging due to the lack of parallel data. We develop a neural approach that includes a hybrid attention-copy mechanism, learns with weak supervisions, and is enhanced with a new content coverage constraint. We conduct experiments in restaurants and sports domains. Results show our approach achieves stronger performance than a range of comparison methods. Our approach balances well between content fidelity and style control given exemplars that match the records to varying degrees.

cs.CL↗

arXiv2018

Measurement of quadratic Terahertz optical nonlinearities using second harmonic lock-in detection

We present a method to measure quadratic Terahertz optical nonlinearities in Terahertz time-domain spectroscopy. We use a rotating linear polarizer (a polarizing chopper) to modulate the amplitude of the incident THz pulse train. We use a phase-sensitive lock-in detection at the fundamental and the second harmonic of the modulation frequency to separate the materials' responses that are linear and quadratic in Terahertz electric field. We demonstrate this method by measuring the quadratic Terahertz Kerr effect in the presence of the much stronger linear electro-optic effect in the (110) GaP crystal. We propose that the method can be used to detect Terahertz second harmonic generation in noncentrosymmetric media in time-domain spectroscopy, with broad potential applications in nonlinear Terahertz photonics and related technology.

physics.app-ph↗

arXiv2017

Giant THz surface plasmon polariton induced by high-index dielectric metasurface

We use computational approaches to explore the role of a high-refractive-index dielectric TiO2 grating with deep subwavelength thickness on InSb as a tunable coupler for THz surface plasmons. We find a series of resonances as the grating couples a normally-incident THz wave to standing surface plasmon waves on both thin and thick InSb layers. In a marked contrast with previously-explored metallic gratings, we observe the emergence of a much stronger additional resonance. The mechanism of this giant plasmonic resonance is well interpreted by the dispersion of surface plasmon excited in the air\TiO2\InSb trilayer system. We demonstrate that both the frequency and the intensity of the giant resonance can be tuned by varying dielectric grating parameters, providing more flexible tunability than metallic gratings. The phase and amplitude of the normally-incident THz wave are spatially modulated by the dielectric grating to optimize the surface plasmon excitation. The giant surface plasmon resonance gives rise to strong enhancement of the electric field above the grating structure, which can be useful in sensing and spectroscopy applications.

physics.app-ph↗

arXiv2016

Thin InSb layers with metallic gratings: a novel platform for spectrally-selective THz plasmonic sensing

We present a computational study of terahertz optical properties of a grating-coupled plasmonic structure based on micrometer-thin InSb layers. We find two strong absorption resonances that we interpret as standing surface plasmon modes and investigate their dispersion relations, dependence on InSb thickness, and the spatial distribution of the electric field. The observed surface plasmon modes are well described by a simple theory of the air/InSb/air trilayer. The plasmonic response of the grating/InSb structure is highly sensitive to the dielectric environment and the presence of an analyte (e.g., lactose) at the InSb interface, which is promising for terahertz plasmonic sensor applications. We determine the sensor sensitivity to be 7200 nm per refractive index unit (or 0.06 THz per refractive index unit). The lower surface plasmon mode also exhibits a splitting when tuned in resonance with the vibrational mode of lactose at 1.37 THz. We propose that such interaction between surface plasmon and vibrational modes can be used as the basis for a new sensing modality that allows the detection of terahertz vibrational fingerprints of an analyte.

cond-mat.mes-hall↗

arXiv2015

Giant negative thermal expansion covering room temperature in nanocrystalline GaNxMn3

Materials with negative thermal expansion (NTE), which contract upon heating, are of great interest both technically and fundamentally. Here, we report giant NTE covering room temperature in mechanically milled antiperovksite GaNxMn3 compounds. The micrograin GaNxMn3 exhibits a large volume contraction at the antiferromagnetic (AFM) to paramagnetic (PM) (AFM-PM) transition within a temperature window ({\Delta}T) of only a few kelvins. The grain size reduces to ~ 30 nm after slight milling, while {\Delta}T is broadened to 50K. The corresponding coefficient of linear thermal expansion ({\alpha}) reaches ~ -70 ppm/K, which is almost two times larger than those obtained in chemically doped antiperovskite compounds. Further reducing grain size to ~ 10 nm, {\Delta}T exceeds 100 K and {\alpha} remains as large as -30 ppm/K (-21 ppm/K) for x = 1.0 (x = 0.9). Excess atomic displacements together with the reduced structural coherence, revealed by high-energy X-ray pair distribution functions, are suggested to delay the AFM-PM transition. By controlling the grain size via mechanically alloying or grinding, giant NTE may also be achievable in other materials with large lattice contraction due to electronic or magnetic phase transitions.

cond-mat.mtrl-sci↗