arXiv subjects

Canran Xu

Publications and source records attributed to Canran Xu.

14 recordsLinked to original sources

arXiv2025

LESER: Learning to Expand via Search Engine-feedback Reinforcement in e-Commerce

User queries in e-commerce search are often vague, short, and underspecified, making it difficult for retrieval systems to match them accurately against structured product catalogs. This challenge is amplified by the one-to-many nature of user intent, where a single query can imply diverse and competing needs. Existing methods, including neural query expansion and prompting-based LLM approaches, fall short in real-world settings: they struggle to capture nuanced user intent, often generate outputs that violate platform constraints, and rely on workflows that are difficult to scale in production. We propose Learning to Expand via Search Engine-feedback Reinforcement (LESER), a novel framework that fine-tunes a context-aware LLM using real-time search engine feedback as supervision. LESER formulates query expansion as a retrieval optimization task and leverages Group Relative Policy Optimization to learn directly from relevance and coverage metrics. LESER is trained to reason over search results and produce high quality query expansions that align with platform rules and retrieval objectives. We evaluate LESER on large-scale, real-world e-commerce datasets, demonstrating substantial improvements in both offline and online settings. Our results show that LESER not only enhances semantic coverage and retrieval relevance but also delivers measurable gains in user engagement, making it a practical and scalable solution for modern search systems.

cs.IR↗

arXiv2025

Bridging Modality Gaps in e-Commerce Products via Vision-Language Alignment

Item information, such as titles and attributes, is essential for effective user engagement in e-commerce. However, manual or semi-manual entry of structured item specifics often produces inconsistent quality, errors, and slow turnaround, especially for Customer-to-Customer sellers. Generating accurate descriptions directly from item images offers a promising alternative. Existing retrieval-based solutions address some of these issues but often miss fine-grained visual details and struggle with niche or specialized categories. We propose Optimized Preference-Based AI for Listings (OPAL), a framework for generating schema-compliant, high-quality item descriptions from images using a fine-tuned multimodal large language model (MLLM). OPAL addresses key challenges in multimodal e-commerce applications, including bridging modality gaps and capturing detailed contextual information. It introduces two data refinement methods: MLLM-Assisted Conformity Enhancement, which ensures alignment with structured schema requirements, and LLM-Assisted Contextual Understanding, which improves the capture of nuanced and fine-grained information from visual inputs. OPAL uses visual instruction tuning combined with direct preference optimization to fine-tune the MLLM, reducing hallucinations and improving robustness across different backbone architectures. We evaluate OPAL on real-world e-commerce datasets, showing that it consistently outperforms baseline methods in both description quality and schema completion rates. These results demonstrate that OPAL effectively bridges the gap between visual and textual modalities, delivering richer, more accurate, and more consistent item descriptions. This work advances automated listing optimization and supports scalable, high-quality content generation in e-commerce platforms.

cs.IR↗

arXiv2024

Investigating LLM Applications in E-Commerce

The emergence of Large Language Models (LLMs) has revolutionized natural language processing in various applications especially in e-commerce. One crucial step before the application of such LLMs in these fields is to understand and compare the performance in different use cases in such tasks. This paper explored the efficacy of LLMs in the e-commerce domain, focusing on instruction-tuning an open source LLM model with public e-commerce datasets of varying sizes and comparing the performance with the conventional models prevalent in industrial applications. We conducted a comprehensive comparison between LLMs and traditional pre-trained language models across specific tasks intrinsic to the e-commerce domain, namely classification, generation, summarization, and named entity recognition (NER). Furthermore, we examined the effectiveness of the current niche industrial application of very large LLM, using in-context learning, in e-commerce specific tasks. Our findings indicate that few-shot inference with very large LLMs often does not outperform fine-tuning smaller pre-trained models, underscoring the importance of task-specific model optimization.Additionally, we investigated different training methodologies such as single-task training, mixed-task training, and LoRA merging both within domain/tasks and between different tasks. Through rigorous experimentation and analysis, this paper offers valuable insights into the potential effectiveness of LLMs to advance natural language processing capabilities within the e-commerce industry.

cs.CL↗

arXiv2022

CCPrefix: Counterfactual Contrastive Prefix-Tuning for Many-Class Classification

Recently, prefix-tuning was proposed to efficiently adapt pre-trained language models to a broad spectrum of natural language classification tasks. It leverages soft prefix as task-specific indicators and language verbalizers as categorical-label mentions to narrow the formulation gap from pre-training language models. However, when the label space increases considerably (i.e., many-class classification), such a tuning technique suffers from a verbalizer ambiguity problem since the many-class labels are represented by semantic-similar verbalizers in short language phrases. To overcome this, inspired by the human-decision process that the most ambiguous classes would be mulled over for each instance, we propose a brand-new prefix-tuning method, Counterfactual Contrastive Prefix-tuning (CCPrefix), for many-class classification. Basically, an instance-dependent soft prefix, derived from fact-counterfactual pairs in the label space, is leveraged to complement the language verbalizers in many-class classification. We conduct experiments on many-class benchmark datasets in both the fully supervised setting and the few-shot setting, which indicates that our model outperforms former baselines.

cs.CL↗

arXiv2021

Hierarchical Character Tagger for Short Text Spelling Error Correction

State-of-the-art approaches to spelling error correction problem include Transformer-based Seq2Seq models, which require large training sets and suffer from slow inference time; and sequence labeling models based on Transformer encoders like BERT, which involve token-level label space and therefore a large pre-defined vocabulary dictionary. In this paper we present a Hierarchical Character Tagger model, or HCTagger, for short text spelling error correction. We use a pre-trained language model at the character level as a text encoder, and then predict character-level edits to transform the original text into its error-free form with a much smaller label space. For decoding, we propose a hierarchical multi-task approach to alleviate the issue of long-tail label distribution without introducing extra model parameters. Experiments on two public misspelling correction datasets demonstrate that HCTagger is an accurate and much faster approach than many existing models.

cs.CL↗

arXiv2020

Response of a quantum disordered spin system to a local periodic drive

We consider a one-dimensional spin chain system with quenched disorder and in the presence of a local periodic drive. We study the time evolution of the system in the Floquet basis and evaluate the fidelity susceptibility, which is a measure of how a given state changes under a small perturbation, of states to a weak periodic drive. We demonstrate that the statistical properties of the fidelity susceptibility over different disorder realizations can be used to identify two phases of the system: (1) the many-body localized phase, in which the susceptibility exhibits long tails while its average value decreases rapidly as disorder increases; and (2) the ergodic phase, in which the susceptibility distribution is narrow and its average value weakly depends on disorder. This distinction in the average value of the susceptibility between the two phases develops readily for systems with ten or more spins. Therefore, recently built networks of qubits subject to a local drive can simulate dynamics of a system in the many-body localization regime. We also show that the spin accumulation speed is correlated with the fidelity susceptibility and can also be used to distinguish the two phases.

cond-mat.dis-nn↗

arXiv2019

Learning Feature Interactions with Lorentzian Factorization Machine

Learning representations for feature interactions to model user behaviors is critical for recommendation system and click-trough rate (CTR) predictions. Recent advances in this area are empowered by deep learning methods which could learn sophisticated feature interactions and achieve the state-of-the-art result in an end-to-end manner. These approaches require large number of training parameters integrated with the low-level representations, and thus are memory and computational inefficient. In this paper, we propose a new model named "LorentzFM" that can learn feature interactions embedded in a hyperbolic space in which the violation of triangle inequality for Lorentz distances is available. To this end, the learned representation is benefited by the peculiar geometric properties of hyperbolic triangles, and result in a significant reduction in the number of parameters (20\% to 80\%) because all the top deep learning layers are not required. With such a lightweight architecture, LorentzFM achieves comparable and even materially better results than the deep learning methods such as DeepFM, xDeepFM and Deep \& Cross in both recommendation and CTR prediction tasks.

cs.IR↗

arXiv2019

Relation Embedding with Dihedral Group in Knowledge Graph

Link prediction is critical for the application of incomplete knowledge graph (KG) in the downstream tasks. As a family of effective approaches for link predictions, embedding methods try to learn low-rank representations for both entities and relations such that the bilinear form defined therein is a well-behaved scoring function. Despite of their successful performances, existing bilinear forms overlook the modeling of relation compositions, resulting in lacks of interpretability for reasoning on KG. To fulfill this gap, we propose a new model called DihEdral, named after dihedral symmetry group. This new model learns knowledge graph embeddings that can capture relation compositions by nature. Furthermore, our approach models the relation embeddings parametrized by discrete values, thereby decrease the solution space drastically. Our experiments show that DihEdral is able to capture all desired properties such as (skew-) symmetry, inversion and (non-) Abelian composition, and outperforms existing bilinear form based approach and is comparable to or better than deep learning models such as ConvE.

cs.CL↗

arXiv2018

Explainable Reasoning over Knowledge Graphs for Recommendation

Incorporating knowledge graph into recommender systems has attracted increasing attention in recent years. By exploring the interlinks within a knowledge graph, the connectivity between users and items can be discovered as paths, which provide rich and complementary information to user-item interactions. Such connectivity not only reveals the semantics of entities and relations, but also helps to comprehend a user's interest. However, existing efforts have not fully explored this connectivity to infer user preferences, especially in terms of modeling the sequential dependencies within and holistic semantics of a path. In this paper, we contribute a new model named Knowledge-aware Path Recurrent Network (KPRN) to exploit knowledge graph for recommendation. KPRN can generate path representations by composing the semantics of both entities and relations. By leveraging the sequential dependencies within a path, we allow effective reasoning on paths to infer the underlying rationale of a user-item interaction. Furthermore, we design a new weighted pooling operation to discriminate the strengths of different paths in connecting a user with an item, endowing our model with a certain level of explainability. We conduct extensive experiments on two datasets about movie and music, demonstrating significant improvements over state-of-the-art solutions Collaborative Knowledge Base Embedding and Neural Factorization Machine.

cs.IR↗

arXiv2015

Response to a local quench of a system near many body localization transition

We consider a one dimensional spin $1/2$ chain with Heisenberg interaction in a disordered parallel magnetic field. This system is known to exhibit the many body localization (MBL) transition at critical strength of disorder. We analyze the response of the chain when additional perpendicular magnetic field is applied to an individual spin and propose a method for accurate determination of the mobility edge via local spin measurements. We further demonstrate that the exponential decrease of the spin response with the distance between perturbed spin and measured spin can be used to characterize the localization length in the MBL phase.

cond-mat.dis-nn↗

arXiv2015

High-fidelity qubit measurement with a microwave photon counter

High-fidelity, efficient quantum nondemolition readout of quantum bits is integral to the goal of quantum computation. As superconducting circuits approach the requirements of scalable, universal fault tolerance, qubit readout must also meet the demand of simplicity to scale with growing system size. Here we propose a fast, high-fidelity, scalable measurement scheme based on the state-selective ring-up of a cavity followed by photodetection with the recently introduced Josephson photomultiplier (JPM), a current-biased Josephson junction. This scheme maps qubit state information to the binary digital output of the JPM, circumventing the need for room-temperature heterodyne detection and offering the possibility of a cryogenic interface to superconducting digital control circuitry. Numerics show that measurement contrast in excess of 95% is achievable in a measurement time of 140 ns. We discuss perspectives to scale this scheme to enable readout of multiple qubit channels with a single JPM.

quant-ph↗

arXiv2014

Nonadiabatic Dynamics of a Dissipative Two-level System

We study the dynamics of a two-level system described by a slowly varying Hamiltonian and weakly coupled to the Ohmic environment. We follow the Bloch--Redfield perturbative approach to include the effect of the environment on qubit evolution and take into account modification of the spectrum and matrix elements of qubit transitions due to time-dependence of the Hamiltonian. This formalism is applied to two problems. (1) We consider a qubit, or a spin-1/2, in a rotating magnetic field. We show that once the rotation starts, the spin has a component perpendicular to the rotation plane of the field that initially wiggles and eventually settles to the value proportional to the product of angular rotation velocity of the field and the Berry curvature. (2) We re-examine the Landau--Zener transition for a system coupled to environment at arbitrary temperature. We show that as temperature increases, the thermal excitation and relaxation become leading processes responsible for transition between states of the system. We also apply the Lindblad master equations to these two problems and compare results with those obtained from the Bloch--Redfield equations.

quant-ph↗

arXiv2013

Full Counting Statistics of Photons Emitted by Double Quantum Dot

We analyze the full counting statistics of photons emitted by a double quantum dot (DQD) coupled to a high-quality microwave resonator by electric dipole interaction. We show that at the resonant condition between the energy splitting of the DQD and the photon energy in the resonator, photon statistics exhibits both a sub-Poissonian distribution and antibunching. In the ideal case, when the system decoherence stems only from photodetection, the photon noise is reduced below one-half of the noise for the Poisson distribution and is consistent with current noise. The photon distribution remains sub-Poissonian even at moderate decoherence in the DQD. We demonstrate that Josephson junction based photomultipliers can be used to experimentally assess statistics of emitted photons.

cond-mat.mes-hall↗

arXiv2013

Quantum Photovoltaic Effect in Double Quantum Dots

We analyze the photovoltaic current through a double quantum dot system coupled to a high-quality driven microwave resonator. The conversion of photons in the resonator to electronic excitations produces a current flow even at zero bias across the leads of the double quantum dot system. We demonstrate that due to the quantum nature of the electromagnetic field in the resonator, the photovoltaic current exhibits a double peak dependence on the frequency $ω$ of an external microwave source. The distance between the peaks is determined by the strength of interaction between photons in the resonator and electrons in the double quantum dot. The double peak structure disappears as strengths of relaxation processes increases, recovering a simple classical condition for maximal current when the microwave frequency is equal to the resonator frequency.

cond-mat.mes-hall↗