arXiv subjects

Wang Xu

Publications and source records attributed to Wang Xu.

33 records · Page 2Linked to original sources

arXiv2025

On the Converse of Pr\'{e}kopa's Theorem and Berndtsson's Theorem

Given a continuous function $\phi$ defined on a domain $\Omega\subset\mathbb{R}^m\times\mathbb{R}^n$, we show that if a Pr\'ekopa-type result holds for $\phi+\psi$ for any non-negative convex function $\psi$ on $\Omega$, then $\phi$ must be a convex function. Additionally, if the projection of $\Omega$ onto $\mathbb{R}^m$ is convex, then $\overline{\Omega}$ is also convex. This provides a converse of Pr\'ekopa's theorem from convex analysis. We also establish analogous results for Berndtsson's theorem on the plurisubharmonic variation of Bergman kernels, showing that the plurisubharmonicity of weight functions and the pseudoconvexity of domains are necessary conditions in some sense.

math.CV↗

arXiv2024

Enabling Real-Time Conversations with Minimal Training Costs

Large language models (LLMs) have demonstrated the ability to improve human efficiency through conversational interactions. Conventional LLM-powered dialogue systems, operating on a turn-based paradigm, preclude real-time interaction during response generation. To address this limitation, researchers have proposed duplex models. These models can dynamically adapt to user input, facilitating real-time interactive feedback. However, these methods typically require substantial computational resources to acquire the ability. To reduce overhead, this paper presents a new duplex decoding approach that enhances LLMs with duplex ability, requiring minimal additional training. Specifically, our method employs parallel decoding of queries and responses in conversations, effectively implementing a channel-division-multiplexing decoding strategy. Experimental results indicate that our proposed method significantly enhances the naturalness and human-likeness of user-AI interactions with minimal training costs.

cs.CL↗

arXiv2024

Ouroboros: Generating Longer Drafts Phrase by Phrase for Faster Speculative Decoding

Speculative decoding is a widely used method that accelerates the generation process of large language models (LLMs) with no compromise in model performance. It achieves this goal by using an existing smaller model for drafting and then employing the target LLM to verify the draft in a low-cost parallel manner. Under such a drafting-verification framework, drafting efficiency has become a bottleneck in the final speedup of speculative decoding. Therefore, generating longer drafts at less cost can lead to better decoding speedup. To achieve this, we introduce Ouroboros, which can generate draft phrases to parallelize the drafting process and meanwhile lengthen drafts in a training-free manner. The experimental results on various typical text generation tasks show that Ouroboros can achieve speedups of up to $2.8\times$ over speculative decoding and $3.9\times$ over vanilla decoding, without fine-tuning draft and target models. The source code of Ouroboros is available at https://github.com/thunlp/Ouroboros.

cs.CL↗

arXiv2024

LoRA-drop: Efficient LoRA Parameter Pruning based on Output Evaluation

Low-Rank Adaptation (LoRA) is currently the most commonly used Parameter-efficient fine-tuning (PEFT) method, it introduces auxiliary parameters for each layer to fine-tune the pre-trained model under limited computing resources. However, it still faces resource consumption challenges during training when scaling up to larger models. Most previous studies have tackled this issue by using pruning techniques, which involve removing LoRA parameters deemed unimportant. Nonetheless, these efforts only analyze LoRA parameter features to evaluate their importance, such as parameter count, size, and gradient. In fact, the output of LoRA (product of LoRA parameter and hidden state), directly impacts the final results. Preliminary experiments indicate that a fraction of LoRA elements possesses significantly high output values, substantially influencing the layer output. Motivated by the observation, we propose LoRA-drop. Concretely, LoRA-drop evaluates the importance of LoRA based on the LoRA output. Then we retain LoRA for important layers and the other layers share the same LoRA. We conduct abundant experiments with models of different scales on NLU and NLG tasks. Results demonstrate that LoRA-drop can achieve performance comparable to full fine-tuning and LoRA, while retaining 50\% of the LoRA parameters on average.

cs.LG↗

arXiv2023

Chinese Fine-Grained Financial Sentiment Analysis with Large Language Models

Entity-level fine-grained sentiment analysis in the financial domain is a crucial subtask of sentiment analysis and currently faces numerous challenges. The primary challenge stems from the lack of high-quality and large-scale annotated corpora specifically designed for financial text sentiment analysis, which in turn limits the availability of data necessary for developing effective text processing techniques. Recent advancements in large language models (LLMs) have yielded remarkable performance in natural language processing tasks, primarily centered around language pattern matching. In this paper, we propose a novel and extensive Chinese fine-grained financial sentiment analysis dataset, FinChina SA, for enterprise early warning. We thoroughly evaluate and experiment with well-known existing open-source LLMs using our dataset. We firmly believe that our dataset will serve as a valuable resource to advance the exploration of real-world financial sentiment analysis tasks, which should be the focus of future research. The FinChina SA dataset is publicly available at https://github.com/YerayL/FinChina-SA

cs.CL↗

arXiv2022

Characterizations of Griffiths Positivity, Pluriharmonicity and Flatness

Deng-Ning-Wang-Zhou showed that a Hermitian holomorphic vector bundle is Griffiths semi-positive if it satisfies the optimal $L^2$-extension condition. As a generalization, we present a quantitative characterization of Griffiths positivity in terms of certain $L^2$-extension conditions. We also show that a $\mathbb{R}$-valued measurable function is pluriharmonic if and only if it satisfies the equality part of the optimal $L^p$-extension condition. This answers a conjecture of Inayama affirmatively. Moreover, the flatness of a possibly singular Hermitian metric is also equivalent to the equality part of the optimal $L^p$-extension condition.

math.CV↗

arXiv2022

Document-Level Relation Extraction with Sentences Importance Estimation and Focusing

Document-level relation extraction (DocRE) aims to determine the relation between two entities from a document of multiple sentences. Recent studies typically represent the entire document by sequence- or graph-based models to predict the relations of all entity pairs. However, we find that such a model is not robust and exhibits bizarre behaviors: it predicts correctly when an entire test document is fed as input, but errs when non-evidence sentences are removed. To this end, we propose a Sentence Importance Estimation and Focusing (SIEF) framework for DocRE, where we design a sentence importance score and a sentence focusing loss, encouraging DocRE models to focus on evidence sentences. Experimental results on two domains show that our SIEF not only improves overall performance, but also makes DocRE models more robust. Moreover, SIEF is a general framework, shown to be effective when combined with a variety of base DocRE models.

cs.CL↗

arXiv2022

Optimal $L^2$ Extensions of Openness Type

We study the following optimal $L^2$ extension problem of openness type: given a complex manifold $M$, a closed subvariety $S\subset M$ and a holomorphic vector bundle $E\rightarrow M$, for any $L^2$ holomorphic section $f$ defined on some open neighborhood $U$ of $S$, find an $L^2$ holomorphic section $F$ on $M$ such that $F|_S = f|_S$, and the $L^2$ norm of $F$ on $M$ is optimally controlled by the $L^2$ norm of $f$ on $U$. Answering the above problem, we prove an optimal $L^2$ extension theorem of openness type on weakly pseudoconvex K\"ahler manifolds, which generalizes a couple of known results on such a problem. Moreover, we prove a product property for certain minimal $L^2$ extensions and give an alternative proof to a version of the above $L^2$ extension theorem. We also present some applications to the usual optimal $L^2$ extension problem and the equality part of Suita's conjecture.

math.CV↗

arXiv2021

Discriminative Reasoning for Document-level Relation Extraction

Document-level relation extraction (DocRE) models generally use graph networks to implicitly model the reasoning skill (i.e., pattern recognition, logical reasoning, coreference reasoning, etc.) related to the relation between one entity pair in a document. In this paper, we propose a novel discriminative reasoning framework to explicitly model the paths of these reasoning skills between each entity pair in this document. Thus, a discriminative reasoning network is designed to estimate the relation probability distribution of different reasoning paths based on the constructed graph and vectorized document contexts for each entity pair, thereby recognizing their relation. Experimental results show that our method outperforms the previous state-of-the-art performance on the large-scale DocRE dataset. The code is publicly available at https://github.com/xwjim/DRN.

cs.CL↗

arXiv2020

Document-Level Relation Extraction with Reconstruction

In document-level relation extraction (DocRE), graph structure is generally used to encode relation information in the input document to classify the relation category between each entity pair, and has greatly advanced the DocRE task over the past several years. However, the learned graph representation universally models relation information between all entity pairs regardless of whether there are relationships between these entity pairs. Thus, those entity pairs without relationships disperse the attention of the encoder-classifier DocRE for ones with relationships, which may further hind the improvement of DocRE. To alleviate this issue, we propose a novel encoder-classifier-reconstructor model for DocRE. The reconstructor manages to reconstruct the ground-truth path dependencies from the graph representation, to ensure that the proposed DocRE model pays more attention to encode entity pairs with relationships in the training. Furthermore, the reconstructor is regarded as a relationship indicator to assist relation classification in the inference, which can further improve the performance of DocRE model. Experimental results on a large-scale DocRE dataset show that the proposed model can significantly improve the accuracy of relation extraction on a strong heterogeneous graph-based baseline.

cs.CL↗

arXiv2013

Test of the prototype of electron detector for LHAASO project using cosmic rays

LHAASO project is to be built in south-west China, which use an array of 5137 election detectors for the measurement of the incident electrons arriving at the detector plane. For the quality control of the big quantity of electron detectors, a cosmic ray hodoscope with two-dimensional spacial sensitivity and good time resolution has been developed. The first prototype of electron detector is tested with the hodoscope and the performance of the detector is validated to be consistent with the design.

physics.ins-det↗

arXiv2008

An Experiment of Research-Oriented Teaching/Learning

We introduce our experiment of research-oriented teaching mainly in Nanjing University and Tsinghua University, China. The great population and enrollment in China makes it worth to concern. It lasts 20 years and involves thousands of students and hundreds of instructors, consultant experts. We tried many characteristic styles such as integrated teaching and case analysis, open resources, interactive mode, course paper program, elite solutions and so on. The research on the contents is also placed on the agenda. Many students joined research works that lead to PRL, APL, Nature, Science, and Cell papers. To impart colleagues the essence we offered some examples in every session. We declare the accomplishment of the experiment through this paper and new project is programming.

physics.ed-ph↗

arXiv2007

Backward Compton Scattering and QED with Noncommutative Plane in the Strong Uniform Magnetic Field

In the strong uniform magnetic field, the noncommutative plane (NCP) caused by the lowest Landau level (LLL) effect, and QED with NCP (QED-NCP) are studied. Being similar to the condensed matter theory of quantum Hall effect, an effective filling factor $f(B)$ is introduced to character the possibility that the electrons stay on the LLL. The analytic and numerical results of the differential cross section for the process of backward Compton scattering in the accelerator with unpolarized or polarized initial photons are calculated. The existing data of BL38B2 in Spring-8 have been analyzed roughly and compared with the numerical predictions primitively. We propose a precise measurement of the differential cross sections of backward Compton scattering in a strong perpendicular magnetic field, which may lead to reveal the effects of QED-NCP.

hep-ph↗

arXiv2006

Backward Compton Scattering in Strong Uniform Magnetic Field

In strong uniform magnetic field, the vacuum Non-Commutative Plane (NCP) caused by the lowest Landau level(LLL) effect and the QED with NCP (QED-NCP) are studied. Being similar to the theory of Quantum Hall effect, an effective filling factor $f(B)$ is introduced to character the possibility that the electrons stays on LLL. The backward Compton scattering amplitudes of QED-NCP are derived, and the differential cross sections for the process with polarized initial electrons and photons are calculated. The existing Spring-8's data has been analyzed primitively and some hints for QED-NCP effects are shown. We propose to precisely measure the differential cross sections of the backward Compton scattering in perpendicular magnetic field experimentally, which may lead to reveal the effects of QED-NCP. PACS number: 12.20.Ds; 11.10.Nx; 71.70.Di; 73.43.Fj.

hep-ph↗

arXiv2004

High Resolution Search for Exotic Pentaquark $Θ^{++}$, and $Θ^+$ at Jefferson Lab

Recent evidence for the existence of pentaquark $Θ^+$ particle from several experiments at several different laboratories around the world has caused great excitement and raised many unanswered questions. We discuss a new proposal on a high resolution experiment searching for pentaquark states in Hall C at Jefferson Lab using an untagged bremsstrahlung photon beam employing both the hydrogen and the deuterium targets by studying the following processes: $γp \to Θ^{++} K^-$, and $γn \to Θ^+ K^-$. This new experiment will significantly improve our current knowledge of the mass, the width of the $Θ^+$ particle if it is confirmed and provide unambiguous evidence for the existence or non-existence of the $Θ^{++}$ particle from the γp \to Θ^{++} K^-$ process.

nucl-ex↗