arXiv subjects

Danyang Liu

Publications and source records attributed to Danyang Liu.

18 recordsLinked to original sources

arXiv2026

Plexciton-mediated Raman scattering in strongly coupled systems

A series of experimental results demonstrate a distinctive Raman response in plasmon-exciton coupled systems. The enhancement of Raman scattering varies for different phonon modes. We describe the microscopic dynamical process of this Raman scattering using quantum many-body theory. Unlike conventional Raman scattering involving electron-phonon interactions, the process in plasmon-exciton coupled systems is characterized by inelastic scattering between phonons and plasmon-exciton polaritons-formed through the coupling of plasmons and excitons-acting as intermediate states. We derive analytical expressions for the Raman intensity and enhancement factors for various phonon modes, which show excellent agreement with experimental data. Furthermore, experimental fittings indicate a substantial disparity in the linewidths of the upper and lower polariton branches, for which we provide a comprehensive theoretical explanation. Based on linear response theory, we propose a microscopic mechanism for the formation of plasmon-exciton polaritons, enabling the analytical calculation of their dispersions and linewidths. This approach naturally accounts for the significantly asymmetry observed in the linewidths of the upper and lower polariton branches. By characterizing the polariton-phonon scattering process at the quantum level, we reveal the fundamental physical mechanism driving polariton-enhanced Raman scattering. Our work establishes a universal framework for describing the dynamical evolution of coupled systems, providing a versatile paradigm for exploring the interactions between plasmons and other quasiparticles.

cond-mat.mes-hall↗

arXiv2026

Stop Overthinking: Unlocking Efficient Listwise Reranking with Minimal Reasoning

Listwise reranking utilizing Large Language Models (LLMs) has achieved state-of-the-art retrieval effectiveness. Recently, reasoning-enhanced models have further pushed these boundaries by employing Chain-of-Thought (CoT) to perform deep comparative analysis of candidate documents. However, this performance gain comes at a prohibitive computational cost, as models often generate thousands of reasoning tokens before producing a final ranking. In this work, we investigate the relationship between reasoning length and ranking quality, revealing an overthinking phenomenon where extended reasoning yields diminishing returns. To address this, we propose a Length-Regularized Self-Distillation framework. We synthesize a dataset by sampling diverse reasoning traces from a teacher model (Rank-K) and applying a Pareto-inspired filter to select traces that achieve high ranking performance with minimal token usage. By fine-tuning on these concise, high-quality rationales, the student model learns to internalize efficient reasoning patterns, effectively pruning redundant deliberation. Experiments on TREC Deep Learning and NeuCLIR benchmarks demonstrate that our method maintains the teacher's effectiveness while reducing inference token consumption by 34%-37% across different retrieval settings, offering a practical solution for deploying reasoning-enhanced rerankers in latency-sensitive applications.

cs.IR↗

arXiv2026

Reversible nanoscale patterning of WTe$_2$ with a scanning tunneling microscope

Manipulating the lattice structure of ferroelectric quantum materials enables their use in low-power electronic devices, including field-effect transistors. WTe$_2$ is a Weyl-semimetal candidate and ferroelectric, both properties arising from the reduced crystal symmetry of its T$_\mathrm{d}$ ground state. The T$_\mathrm{d}$ crystal phase results from a Peierls distortion of the 1T parent structure and an interlayer shift. While experiments in WTe$_2$ have established ferroelectric switching and transient control of the predicted topological phase via ultrafast excitations, persistent electronic changes on the nanometer scale remain elusive. Here, we demonstrate that current pulses applied via scanning tunneling microscopy can both write and erase persistent nanometer-scale patterns on the surface of WTe$_2$. These patterns consist of apparent picometer in-plane and out-of-plane atomic displacements, accompanied by changes to the local density of states. The out-of-plane displacements further modulate the Peierls-like distortion present in WTe$_2$, while the in-plane displacements are indicative of ferroelectric switching. The induced patterns can be repositioned and erased, suggesting a nanoscale handle on the ferroelectric properties of WTe$_2$.

cond-mat.mtrl-sci↗

arXiv2025

Existence of Solutions and Relative Regularity Conditions for Polynomial Vector Optimization Problems

In this paper, we establish the existence of the efficient solutions for polynomial vector optimization problems on a nonempty closed constraint set without any convexity and compactness assumptions. We first introduce the relative regularity conditions for vector optimization problems whose objective functions are a vector polynomial and investigate their properties and characterizations. Moreover, we establish relationships between the relative regularity conditions, Palais-Smale condition, weak Palais-Smale condition, M-tameness and properness with respect to some index set. Under the relative regularity and non-regularity conditions, we establish nonemptiness of the efficient solution sets of the polynomial vector optimization problems respectively. As a by-product, we infer Frank-Wolfe type theorems for a non-convex polynomial vector optimization problem. Finally, we study the local properties and genericity characteristics of the relative regularity conditions.

math.OC↗

arXiv2024

Generating Visual Stories with Grounded and Coreferent Characters

Characters are important in narratives. They move the plot forward, create emotional connections, and embody the story's themes. Visual storytelling methods focus more on the plot and events relating to it, without building the narrative around specific characters. As a result, the generated stories feel generic, with character mentions being absent, vague, or incorrect. To mitigate these issues, we introduce the new task of character-centric story generation and present the first model capable of predicting visual stories with consistently grounded and coreferent character mentions. Our model is finetuned on a new dataset which we build on top of the widely used VIST benchmark. Specifically, we develop an automated pipeline to enrich VIST with visual and textual character coreference chains. We also propose new evaluation metrics to measure the richness of characters and coreference in stories. Experimental results show that our model generates stories with recurring characters which are consistent and coreferent to larger extent compared to baselines and state-of-the-art systems.

cs.CL↗

arXiv2024

Existence of Weak Pareto Efficient Solutions of a Vector Optimization Problem under a Closed Constraint Set

In this paper, we investigate the nonemptiness of weak Pareto efficient solution set for a class of nonsmooth vector optimization problems on a nonempty closed constraint set without any boundedness and convexity assumptions. First, we obtain a new property concerning the nonemptiness of weak Pareto efficient solution sets for these vector optimization problems. Then, under the condition of weak section-boundedness from below, we establish relationships between the notions of properness, Palais-Smale, weak Palais-Smale and M-tameness conditions with respect to some index sets for the restriction of the vector mapping on the constraint set. Finally, we present some new necessary and sufficient conditions for the nonemptiness of the weak Pareto efficient solution set for a general class of nonsmooth vector optimization problems.

math.OC↗

arXiv2024

TrustScore: Reference-Free Evaluation of LLM Response Trustworthiness

Large Language Models (LLMs) have demonstrated impressive capabilities across various domains, prompting a surge in their practical applications. However, concerns have arisen regarding the trustworthiness of LLMs outputs, particularly in closed-book question-answering tasks, where non-experts may struggle to identify inaccuracies due to the absence of contextual or ground truth information. This paper introduces TrustScore, a framework based on the concept of Behavioral Consistency, which evaluates whether an LLMs response aligns with its intrinsic knowledge. Additionally, TrustScore can seamlessly integrate with fact-checking methods, which assesses alignment with external knowledge sources. The experimental results show that TrustScore achieves strong correlations with human judgments, surpassing existing reference-free metrics, and achieving results on par with reference-based metrics.

cs.CL↗

arXiv2023

Visual Storytelling with Question-Answer Plans

Visual storytelling aims to generate compelling narratives from image sequences. Existing models often focus on enhancing the representation of the image sequence, e.g., with external knowledge sources or advanced graph structures. Despite recent progress, the stories are often repetitive, illogical, and lacking in detail. To mitigate these issues, we present a novel framework which integrates visual representations with pretrained language models and planning. Our model translates the image sequence into a visual prefix, a sequence of continuous embeddings which language models can interpret. It also leverages a sequence of question-answer pairs as a blueprint plan for selecting salient visual concepts and determining how they should be assembled into a narrative. Automatic and human evaluation on the VIST benchmark (Huang et al., 2016) demonstrates that blueprint-based models generate stories that are more coherent, interesting, and natural compared to competitive baselines and state-of-the-art systems.

cs.CL↗

arXiv2023

Modeling Spatiotemporal Periodicity and Collaborative Signal for Local-Life Service Recommendation

Online local-life service platforms provide services like nearby daily essentials and food delivery for hundreds of millions of users. Different from other types of recommender systems, local-life service recommendation has the following characteristics: (1) spatiotemporal periodicity, which means a user's preferences for items vary from different locations at different times. (2) spatiotemporal collaborative signal, which indicates similar users have similar preferences at specific locations and times. However, most existing methods either focus on merely the spatiotemporal contexts in sequences, or model the user-item interactions without spatiotemporal contexts in graphs. To address this issue, we design a new method named SPCS in this paper. Specifically, we propose a novel spatiotemporal graph transformer (SGT) layer, which explicitly encodes relative spatiotemporal contexts, and aggregates the information from multi-hop neighbors to unify spatiotemporal periodicity and collaborative signal. With extensive experiments on both public and industrial datasets, this paper validates the state-of-the-art performance of SPCS.

cs.IR↗

arXiv2023

Detecting and Grounding Important Characters in Visual Stories

Characters are essential to the plot of any story. Establishing the characters before writing a story can improve the clarity of the plot and the overall flow of the narrative. However, previous work on visual storytelling tends to focus on detecting objects in images and discovering relationships between them. In this approach, characters are not distinguished from other objects when they are fed into the generation pipeline. The result is a coherent sequence of events rather than a character-centric story. In order to address this limitation, we introduce the VIST-Character dataset, which provides rich character-centric annotations, including visual and textual co-reference chains and importance ratings for characters. Based on this dataset, we propose two new tasks: important character detection and character grounding in visual stories. For both tasks, we develop simple, unsupervised models based on distributional similarity and pre-trained vision-and-language models. Our new dataset, together with these models, can serve as the foundation for subsequent work on analysing and generating stories from a character-centric perspective.

cs.CL↗

arXiv2023

Adaptive Resource Allocation for Workflow Containerization on Kubernetes

In a cloud-native era, the Kubernetes-based workflow engine enables workflow containerized execution through the inherent abilities of Kubernetes. However, when encountering continuous workflow requests and unexpected resource request spikes, the engine is limited to the current workflow load information for resource allocation, which lacks the agility and predictability of resource allocation, resulting in over and under-provisioning resources. This mechanism seriously hinders workflow execution efficiency and leads to high resource waste. To overcome these drawbacks, we propose an adaptive resource allocation scheme named ARAS for the Kubernetes-based workflow engines. Considering potential future workflow task requests within the current task pod's lifecycle, the ARAS uses a resource scaling strategy to allocate resources in response to high-concurrency workflow scenarios. The ARAS offers resource discovery, resource evaluation, and allocation functionalities and serves as a key component for our tailored workflow engine (KubeAdaptor). By integrating the ARAS into KubeAdaptor for workflow containerized execution, we demonstrate the practical abilities of KubeAdaptor and the advantages of our ARAS. Compared with the baseline algorithm, experimental evaluation under three distinct workflow arrival patterns shows that ARAS gains time-saving of 9.8% to 40.92% in the average total duration of all workflows, time-saving of 26.4% to 79.86% in the average duration of individual workflow, and an increase of 1% to 16% in CPU and memory resource usage rate.

cs.DC↗

arXiv2022

A versatile platform for graphene nanoribbon synthesis, electronic decoupling, and spin polarized measurements

The on-surface synthesis of nano-graphenes has led the charge in prototyping structures with perspectives beyond silicon-based technology. Following reports of open-shell systems in graphene-nanoribbons (GNR), a flurry of research activities is directed at investigating their magnetic properties with a keen eye for spintronic applications. Although the synthesis of nano-graphenes is usually straightforward on gold, it is difficult to use it for electronic decoupling and spin-polarized measurements. Using a binary alloy Cu3Au(111), we show how to combine the efficient gold-like nano-graphene formation with spin polarization and electronic decoupling known from copper. We prepare copper oxide layers, demonstrate thermally and tip-assisted synthesis of GNR and grow thermally stable magnetic Co islands. We functionalize the tip of a scanning tunneling microscope with carbon-monoxide, nickelocene, or attach Co clusters for high-resolution imaging, magnetic sensing, or spin-polarized measurements. This versatile platform will be a valuable tool in the advanced study of magnetic nano-graphenes.

cond-mat.mes-hall↗

arXiv2022

A Sacrificial Magnet Concept for Field Dependent Surface Science Studies

We demonstrate a straightforward approach to integrating a magnetic field into a low-temperature scanning tunneling microscope (STM) by adhering an NdFeB permanent magnet to a magnetizable sample plate. To render our magnet concept compatible with high-temperature sample cleaning procedures, we make the irreversible demagnetization of the magnet a central part of our preparation cycle. After sacrificing the magnet by heating it above its Curie temperature, we use a transfer tool to attach a new magnet in-situ prior to transferring the sample into the STM. We characterize the magnetic field created by the magnet using the Abrikosov vortex lattice of superconducting NbSe2. Excellent agreement between the distance dependent magnetic fields from experiments and simulations allows us to predict the magnitude and orientation of magnetic flux at any location with respect to the magnet and the sample plate. Our concept is an accessible solution for field-dependent surface science studies that require fields in the range of up to 400 mT and otherwise detrimental heating procedures.

physics.app-ph↗

arXiv2022

Adaptive Sparse Sampling for Quasiparticle Interference Imaging

Quasiparticle interference imaging (QPI) offers insight into the band structure of quantum materials from the Fourier transform of local density of states (LDOS) maps. Their acquisition with a scanning tunneling microscope is traditionally tedious due to the large number of required measurements that may take several days to complete. The recent demonstration of sparse sampling for QPI imaging showed how the effective measurement time could be fundamentally reduced by only sampling a small and random subset of the total LDOS. However, the amount of required sub-sampling to faithfully recover the QPI image remained a recurring question. Here we introduce an adaptive sparse sampling (ASS) approach in which we gradually accumulate sparsely sampled LDOS measurements until a desired quality level is achieved via compressive sensing recovery. The iteratively measured random subset of the LDOS can be interleaved with regular topographic images that are used for image registry and drift correction. These reference topographies also allow to resume interrupted measurements to further improve the QPI quality. Our ASS approach is a convenient extension to quasiparticle interference imaging that should remove further hesitation in the implementation of sparse sampling mapping schemes.

cond-mat.other↗

arXiv2021

Fast spectroscopic mapping of two-dimensional quantum materials

Spectroscopic mapping refers to the massive recording of spectra whilst varying an additional degree of freedom, such as: magnetic field, location, temperature, or charge carrier concentration. As this involves two serial tasks, spectroscopic mapping can become excruciatingly slow. We demonstrate exponentially faster mapping through our combination of sparse sampling and parallel spectroscopy. We exemplify our concept using quasiparticle interference imaging of Au(111) and Bi2Sr2CaCu2O8 (Bi2212), as two well-known model systems. Our method is accessible, straightforward to implement with existing scanning tunneling microscopes, and can be easily extended to enhance gate or field-mapping spectroscopy. In view of a possible four orders of magnitude speed advantage, it is setting the stage to fundamentally promote the discovery of novel quantum materials.

cond-mat.mtrl-sci↗

arXiv2021

Solvability of a Regular Polynomial Vector Optimization Problem without Convexity

In this paper we consider the solvability of a non-convex regular polynomial vector optimization problem on a nonempty closed set. We introduce regularity conditions for the polynomial vector optimization problem and study properties and characterizations of the regularity conditions. Under the regularity conditions, we study nonemptiness and boundedness of the solution sets of the problem. As a consequence, we establish two Frank-Wolfe type theorems for the non-convex polynomial vector optimization problem. Finally, we investigate the solution stability of the non-convex regular polynomial vector optimization problem.

math.OC↗

arXiv2020

KRED: Knowledge-Aware Document Representation for News Recommendations

News articles usually contain knowledge entities such as celebrities or organizations. Important entities in articles carry key messages and help to understand the content in a more direct way. An industrial news recommender system contains various key applications, such as personalized recommendation, item-to-item recommendation, news category classification, news popularity prediction and local news detection. We find that incorporating knowledge entities for better document understanding benefits these applications consistently. However, existing document understanding models either represent news articles without considering knowledge entities (e.g., BERT) or rely on a specific type of text encoding model (e.g., DKN) so that the generalization ability and efficiency is compromised. In this paper, we propose KRED, which is a fast and effective model to enhance arbitrary document representation with a knowledge graph. KRED first enriches entities' embeddings by attentively aggregating information from their neighborhood in the knowledge graph. Then a context embedding layer is applied to annotate the dynamic context of different entities such as frequency, category and position. Finally, an information distillation layer aggregates the entity embeddings under the guidance of the original document representation and transforms the document vector into a new one. We advocate to optimize the model with a multi-task framework, so that different news recommendation applications can be united and useful information can be shared across different tasks. Experiments on a real-world Microsoft News dataset demonstrate that KRED greatly benefits a variety of news recommendation applications.

cs.IR↗

arXiv2019

Collaborative Metric Learning with Memory Network for Multi-Relational Recommender Systems

The success of recommender systems in modern online platforms is inseparable from the accurate capture of users' personal tastes. In everyday life, large amounts of user feedback data are created along with user-item online interactions in a variety of ways, such as browsing, purchasing, and sharing. These multiple types of user feedback provide us with tremendous opportunities to detect individuals' fine-grained preferences. Different from most existing recommender systems that rely on a single type of feedback, we advocate incorporating multiple types of user-item interactions for better recommendations. Based on the observation that the underlying spectrum of user preferences is reflected in various types of interactions with items and can be uncovered by latent relational learning in metric space, we propose a unified neural learning framework, named Multi-Relational Memory Network (MRMN). It can not only model fine-grained user-item relations but also enable us to discriminate between feedback types in terms of the strength and diversity of user preferences. Extensive experiments show that the proposed MRMN model outperforms competitive state-of-the-art algorithms in a wide range of scenarios, including e-commerce, local services, and job recommendations.

cs.LG↗