arXiv subjects

Feng Wei

Publications and source records attributed to Feng Wei.

At least 37 records · Page 2Linked to original sources

arXiv2025

QUILL: Quotation Generation Enhancement of Large Language Models

While Large language models (LLMs) have become excellent writing assistants, they still struggle with quotation generation. This is because they either hallucinate when providing factual quotations or fail to provide quotes that exceed human expectations. To bridge the gap, we systematically study how to evaluate and improve LLMs' performance in quotation generation tasks. We first establish a holistic and automatic evaluation system for quotation generation task, which consists of five criteria each with corresponding automatic metric. To improve the LLMs' quotation generation abilities, we construct a bilingual knowledge base that is broad in scope and rich in dimensions, containing up to 32,022 quotes. Moreover, guided by our critiria, we further design a quotation-specific metric to rerank the retrieved quotations from the knowledge base. Extensive experiments show that our metrics strongly correlate with human preferences. Existing LLMs struggle to generate desired quotes, but our quotation knowledge base and reranking metric help narrow this gap. Our dataset and code are publicly available at https://github.com/GraceXiaoo/QUILL.

cs.CL↗

arXiv2024

MultiLingPoT: Enhancing Mathematical Reasoning with Multilingual Program Fine-tuning

Program-of-Thought (PoT), which aims to use programming language instead of natural language as an intermediate step in reasoning, is an important way for LLMs to solve mathematical problems. Since different programming languages excel in different areas, it is natural to use the most suitable language for solving specific problems. However, current PoT research only focuses on single language PoT, ignoring the differences between different programming languages. Therefore, this paper proposes an multilingual program reasoning method, MultiLingPoT. This method allows the model to answer questions using multiple programming languages by fine-tuning on multilingual data. Additionally, prior and posterior hybrid methods are used to help the model select the most suitable language for each problem. Our experimental results show that the training of MultiLingPoT improves each program's mathematical reasoning by about 2.5\%. Moreover, with proper mixing, the performance of MultiLingPoT can be further improved, achieving a 6\% increase compared to the single-language PoT with the data augmentation.Resources of this paper can be found at https://github.com/Nianqi-Li/MultiLingPoT.

cs.CL↗

arXiv2024

SEGMENT+: Long Text Processing with Short-Context Language Models

There is a growing interest in expanding the input capacity of language models (LMs) across various domains. However, simply increasing the context window does not guarantee robust performance across diverse long-input processing tasks, such as understanding extensive documents and extracting detailed information from lengthy and noisy data. In response, we introduce SEGMENT+, a general framework that enables LMs to handle extended inputs within limited context windows efficiently. SEGMENT+ utilizes structured notes and a filtering module to manage information flow, resulting in a system that is both controllable and interpretable. Our extensive experiments across various model sizes, focusing on long-document question-answering and Needle-in-a-Haystack tasks, demonstrate the effectiveness of SEGMENT+ in improving performance.

cs.CL↗

arXiv2024

The Giant Radio Array for Neutrino Detection (GRAND) Collaboration -- Contributions to the 38th International Cosmic Ray Conference (ICRC 2023)

The Giant Radio Array for Neutrino Detection (GRAND) is an envisioned observatory of ultra-high-energy particles of cosmic origin, with energies in excess of 100 PeV. GRAND uses large surface arrays of autonomous radio-detection units to look for the radio emission from extensive air showers that are triggered by the interaction of ultra-high-energy cosmic rays, gamma rays, and neutrinos in the atmosphere or underground. In particular, for ultra-high-energy neutrinos, the future final phase of GRAND aims to be sensitive enough to discover them in spite of their plausibly tiny flux. Presently, three prototype GRAND radio arrays are in operation: GRANDProto300, in China, GRAND@Auger, in Argentina, and GRAND@Nancay, in France. Their goals are to field-test the design of the radio-detection units, understand the radio background to which they are exposed, and develop tools for diagnostic, data gathering, and data analysis. This list of contributions to the 38th International Cosmic Ray Conference (ICRC 2023) presents an overview of GRAND, in its present and future incarnations, and a look at the first data collected by GRANDProto13, the first phase of GRANDProto300.

hep-ex↗

arXiv2024

The Giant Radio Array for Neutrino Detection (GRAND) Collaboration -- Contributions to the 10th International Workshop on Acoustic and Radio EeV Neutrino Detection Activities (ARENA 2024)

This is an index of the contributions by the Giant Radio Array for Neutrino Detection (GRAND) Collaboration to the 10th International Workshop on Acoustic and Radio EeV Neutrino Detection Activities (ARENA 2024, University of Chicago, June 11-14, 2024). The contributions include an overview of GRAND in its present and future incarnations, methods of radio-detection that are being developed for them, and ongoing joint work between the GRAND and BEACON experiments.

astro-ph.IM↗

arXiv2024

Teaching Large Language Models to Express Knowledge Boundary from Their Own Signals

Large language models (LLMs) have achieved great success, but their occasional content fabrication, or hallucination, limits their practical application. Hallucination arises because LLMs struggle to admit ignorance due to inadequate training on knowledge boundaries. We call it a limitation of LLMs that they can not accurately express their knowledge boundary, answering questions they know while admitting ignorance to questions they do not know. In this paper, we aim to teach LLMs to recognize and express their knowledge boundary, so they can reduce hallucinations caused by fabricating when they do not know. We propose CoKE, which first probes LLMs' knowledge boundary via internal confidence given a set of questions, and then leverages the probing results to elicit the expression of the knowledge boundary. Extensive experiments show CoKE helps LLMs express knowledge boundaries, answering known questions while declining unknown ones, significantly improving in-domain and out-of-domain performance.

cs.CL↗

arXiv2022

Semi-Supervised Clustering with Contrastive Learning for Discovering New Intents

Most dialogue systems in real world rely on predefined intents and answers for QA service, so discovering potential intents from large corpus previously is really important for building such dialogue services. Considering that most scenarios have few intents known already and most intents waiting to be discovered, we focus on semi-supervised text clustering and try to make the proposed method benefit from labeled samples for better overall clustering performance. In this paper, we propose Deep Contrastive Semi-supervised Clustering (DCSC), which aims to cluster text samples in a semi-supervised way and provide grouped intents to operation staff. To make DCSC fully utilize the limited known intents, we propose a two-stage training procedure for DCSC, in which DCSC will be trained on both labeled samples and unlabeled samples, and achieve better text representation and clustering performance. We conduct experiments on two public datasets to compare our model with several popular methods, and the results show DCSC achieve best performance across all datasets and circumstances, indicating the effect of the improvements in our work.

cs.LG↗

arXiv2021

Multivariable $(φ,Γ)$-modules and Representations of Products of Galois Groups: The Case of Imperfect Residue Field

Let $K$ be a complete discretely valued field with mixed characteristic $(0, p)$ and imperfect residue field $k_K$. Let $Δ$ be a finite set. We construct an equivalence of categories between finite dimensional $\Bbb{F}_p$-representations of the product of $Δ$ copies of the absolute Galois group of $K$ and multivariable \' etale $(φ, Γ)$-modules over a multivariable Laurent series ring over $k_K$.

math.NT↗

arXiv2021

Nonlinear Lie-Type Derivations of finitary Incidence Algebras and Related Topics

This is a continuation of our earlier works \cite{KhrypchenkoWei, Yang20211, Yang20212} with respect to (non-)linear Lie-type derivations of finitary incidence algebras. Let $X$ be a pre-ordered set, $\mathcal{R}$ be a $2$-torsionfree and $(n-1)$-torsionfree commutative ring with identity, where $n\geq 2$ is an integer. Let $FI(X,\mathcal{R})$ be the finitary incidence algebra of $X$ over $\mathcal{R}$. In this paper, a complete clarification is obtained for the structure of nonlinear Lie-type derivations of $FI(X,\mathcal{R})$. We introduce a new class of derivations on $FI(X,\mathcal{R})$ named inner-like derivations, and prove that each nonlinear Lie $n$-derivation on $FI(X,\mathcal{R})$ is the sum of an inner-like derivation, a transitive induced derivation and a quasi-additive induced Lie $n$-derivation. Furthermore, if $X$ is finite, we show that a quasi-additive induced Lie $n$-derivation can be expressed as the sum of an additive induced Lie derivation and a central-valued map annihilating all $(n-1)$-th commutators. We also provide a sufficient and necessary condition such that every nonlinear Lie $n$-derivation of $FI(X,\mathcal{R})$ is of proper form. Some related topics for further research are proposed in the last section of this article.

math.RA↗

arXiv2021

Thermomechanical properties of polypropylene and styrene-ethylene-butylene-styrene blends: a molecular simulation and experimental study

Polymer blends consisting of two or more polymers are important for a wide variety of industries and processes, but, the precise mechanism of their thermomechanical behaviour is incompletely understood. In order to understand clearly, it is essential to determine the miscibility and interactions between the components in polymer blend and its macroscopic thermomechanical properties. In this study, we performed experiments on SEBS and isotactic PP blends (SP) as well as molecular dynamics simulations, aiming to know the role played by molecular interactions on the thermomechanical properties. To investigate the glass transition temperature (Tg) of SEBS, PP and their blends at different ratio, the unit cell of the polymer molecular structure of each was established. The LAMMPS molecular dynamics method was used to predict the density, specific volume, free volume, enthalpy, kinetic energy, potential energy and bond energy. The (Tg) s of the SEBS, PP and SP blends were predicted by analysing these properties. Interestingly, the simulated values of the Tg of SEBS, PP and their blends showed good agreement with our experimental results obtained from dynamic mechanical analysis (DMA). This technique used in this work can be used in studying glass transition of other complex polymer blends.

cond-mat.soft↗

arXiv2020

K-Commuting Mappings of Generalized Matrix Algebras

In this paper we will study $k$-commuting mappings of generalized matrix algebras. The general form of arbitrary $k$-commuting mapping of a generalized matrix algebra is determined. It is shown that under mild assumptions, every $k$-commuting mapping of a generalized matrix algebra takes a certain form which is said to be proper. A number of applications related to $k$-commuting mappings are presented. These results extend the existing works of Cheung, Du and Wang \cite{Cheung2, DuWang1} to the case of generalized matrix

math.RA↗

arXiv2020

Lie Biderivations on Triangular Algebras

Let $\mathcal{T}$ be a triangular algebra over a commutative ring $\mathcal{R}$ and $φ: \mathcal{T} \times \mathcal{T}\longrightarrow \mathcal{T}$ be an arbitrary Lie biderivation of $\mathcal{T}$. We will address the question of describing the form of $φ$ in the current work. It is shown that under certain mild assumptions, $φ$ is the sum of an inner biderivation and an extremal biderivation and a some central bilinear mapping. Our results is immediately applied to block upper triangular algebras and Hilbert space nest algebras .

math.RA↗

arXiv2020

Twitter Bot Detection Using Bidirectional Long Short-term Memory Neural Networks and Word Embeddings

Twitter is a web application playing dual roles of online social networking and micro-blogging. The popularity and open structure of Twitter have attracted a large number of automated programs, known as bots. Legitimate bots generate a large amount of benign contextual content, i.e., tweets delivering news and updating feeds, while malicious bots spread spam or malicious contents. To assist human users in identifying who they are interacting with, this paper focuses on the classification of human and spambot accounts on Twitter, by employing recurrent neural networks, specifically bidirectional Long Short-term Memory (BiLSTM), to efficiently capture features across tweets. To the best of our knowledge, our work is the first that develops a recurrent neural model with word embeddings to distinguish Twitter bots from human accounts, that requires no prior knowledge or assumption about users' profiles, friendship networks, or historical behavior on the target account. Moreover, our model does not require any handcrafted features. The preliminary simulation results are very encouraging. Experiments on the cresci-2017 dataset show that our approach can achieve competitive performance compared with existing state-of-the-art bot detection systems.

cs.CR↗

arXiv2020

Multiplicative Lie-type derivations on alternative rings

Let $\R$ be an alternative ring containing a nontrivial idempotent and $\D$ be a multiplicative Lie-type derivation from $\R$ into itself. Under certain assumptions on $\R$, we prove that $\D$ is almost additive. Let $p_n(x_1, x_2, \cdots, x_n)$ be the $(n-1)$-th commutator defined by $n$ indeterminates $x_1, \cdots, x_n$. If $\R$ is a unital alternative ring with a nontrivial idempotent and is $\{2,3,n-1,n-3\}$-torsion free, it is shown under certain condition of $\R$ and $\D$, that $\D=δ+τ$, where $δ$ is a derivation and $τ\colon\R\longrightarrow{\mathcal Z}(\R)$ such that $τ(p_n(a_1,\ldots,a_n))=0$ for all $a_1,\ldots,a_n\in\R$.

math.RA↗

arXiv2019

Non-existence of non-trivial normal elements in the Iwasawa Algebra of Chevalley groups

For a prime $p>2$, let $G$ be a semi-simple, simply connected, split Chevalley group over $\mathbb{Z}_p$, $G(1)$ be the first congruence kernel of $G$ and $Ω_{G(1)}$ be the mod-$p$ Iwasawa algebra defined over the finite field $\mathbb{F}_p$. Ardakov, Wei, Zhang have shown that if $p$ is a "nice prime " ($p \geq 5$ and $p \nmid n+1$ if the Lie algebra of $G(1)$ is of type $A_n$), then every non-zero normal element in $Ω_{G(1)}$ is a unit. Furthermore, they conjecture in their paper that their nice prime condition is superfluous. The main goal of this article is to provide an entirely new proof of Ardakov, Wei and Zhang's result using explicit presentation of Iwasawa algebra developed by the second author of this article and thus eliminating the nice prime condition, therefore proving their conjecture.

math.NT↗

arXiv2019

Dual-FOFE-net Neural Models for Entity Linking with PageRank

This paper presents a simple and computationally efficient approach for entity linking (EL), compared with recurrent neural networks (RNNs) or convolutional neural networks (CNNs), by making use of feedforward neural networks (FFNNs) and the recent dual fixed-size ordinally forgetting encoding (dual-FOFE) method to fully encode the sentence fragment and its left/right contexts into a fixed-size representation. Furthermore, in this work, we propose to incorporate PageRank based distillation in our candidate generation module. Our neural linking models consist of three parts: a PageRank based candidate generation module, a dual-FOFE-net neural ranking model and a simple NIL entity clustering system. Experimental results have shown that our proposed neural linking models achieved higher EL accuracy than state-of-the-art models on the TAC2016 task dataset over the baseline system, without requiring any in-house data or complicated handcrafted features. Moreover, it achieves a competitive accuracy on the TAC2017 task dataset.

cs.CL↗

arXiv2019

Lie-type Derivations of Finitary Incidence Algebras

Let $P$ be an arbitrary partially ordered set, $R$ a commutative ring with identity and $FI(P,R)$ the finitary incidence algebra of $P$ over $R$. Under some natural assumption on $R$, we prove that each Lie-type derivation of $FI(P,R)$ is proper.

math.RA↗

arXiv2018

Higher Derivations of Finitary Incidence Algebras

Let $P$ be a partially ordered set, $R$ a commutative unital ring and $FI(P,R)$ the finitary incidence algebra of $P$ over $R$. We prove that each $R$-linear higher derivation of $FI(P,R)$ decomposes into the product of an inner higher derivation of $FI(P,R)$ and the higher derivation of $FI(P,R)$ induced by a higher transitive map on the set of segments of $P$.

math.RA↗