id
int64
1
361k
content
stringlengths
1
23.3M
title
stringlengths
1
271
appendix
bool
1 class
paper_arxiv_id
stringlengths
10
13
section_in_paper_id
float64
1
294
95,360
\label{sec:conclusion} In this work, we present DiffSpectra, a novel spectrum-conditioned diffusion framework for molecular structure elucidation. By integrating a continuous-time variational diffusion model with a Transformer-based molecular graph backbone, DiffSpectra is capable of generating molecular graphs with c...
Conclusion
false
2507.06853
3
95,361
\label{sec:method} \subsection{Diffusion Model for Molecular Structure Elucidation} We adopt a continuous-time diffusion probabilistic model to generate molecular graphs in a joint space that integrates 2D topology and 3D geometry. Each molecule is represented as a graph $\mathcal{G} = (\mathbf{H}, \mathbf{A}, \math...
Methods
false
2507.06853
4
95,362
\vspace{2pt}
Appendix
false
2507.06853
5
95,363
This appendix derives the reverse sampling formula for our continuous-time diffusion model, as applied to molecular graph generation. Our objective is to obtain an expression for sampling a less-noisy graph state $\cG_s$\cG_s from a more-noisy state $\cG_t = (\bH_t, \bA_t, \bX_t)$\cG_t = (\bH_t, \bA_t, \bX_t), given ...
Derivation of the Reverse Sampling Process
false
2507.06853
6
95,364
We conduct experiments on the QM9S dataset~\cite{DetaNet,MolSpectra}, which augments the original QM9~\citep{QM9} molecular dataset with simulated spectra. For each molecule, we extract its structure and corresponding IR, Raman, and UV-Vis spectra, forming a multi-modal spectral condition for our structure elucidati...
Datasets
false
2507.06853
7
95,365
Aiming to generating chemically valid and complete molecules, modeling accurate molecular distributions, and achieving achieve accurate sturcture elucidation from spectra, we elaborately design a series of evaluation metrics to reflect the generation quality. Specifically, we introduce two categories of evaluation ...
Metrics
false
2507.06853
8
95,359
\label{sec:results} \subsection{Overview of DiffSpectra Framework} To achieve accurate molecular structure elucidation from molecular spectra, we propose \themodel, a spectrum-conditioned diffusion framework that generates molecular structures guided by spectral information. The overall paradigm is illustrated in \c...
Results
false
2507.06853
2
95,369
Now we want to define the notion of derivability for $R$R-expressions of arbitrary $R$R-degree. Therefore, we define a calculus in which not only sequents but $R$R-expressions of arbitrary degree are allowed. The following structural rules hold: \begin{center} \vspace{0.4cm} \quad \infer[RF]{S\Rightarrow S}{}\hspac...
Higher-Order Derivability
false
2507.06854
4
95,366
From the viewpoint of what would nowadays be called `(bilateralist) proof-theoretic semantics' Franz von Kutschera provided two very relevant and closely related papers \cite{Kutschera1968,Kutschera1969}, in which he proves functional completeness of, respectively, intuitionistic logic and what he calls `direct logic'...
Introduction
false
2507.06854
1
95,367
There has been a recent increase in attention with respect to studying connexive logics, resulting in a number of papers published on the subject.\footnote{See, e.g., the contributions in the special issue of \emph{Studia Logica} on \emph{Frontiers of Connexive Logic} \cite{OmoriWansingSI}.} The general idea of conne...
The motivation: A contradictory logic that is functionally complete
false
2507.06854
2
3
\label{TDA1D} In this section, we sketch the basic ideas of the rapidly expanding field of topological data analysis by considering a simple application in one dimension that simultaneously advances the state of the art in the fundamental area of nonparametric statistical estimation and avoids much of the technical ba...
Topological data analysis and unsupervised learning in one dimension
false
2008.03299
3
5
\label{Conclusion} We have only scratched the surface of topological techniques that can be fruitfully applied to problems in the cyber domain. Discrete Morse theory \cite{Scoville_2019}, the algebraic topology of finite topological spaces \cite{barmak2011algebraic}, and connections between simplicial complexes and pa...
Conclusion
false
2008.03299
5
16,060
\label{intro} Transportation infrastructure is essential to modern society, forming the backbone of economic development by enabling the efficient movement of people and goods. Among these infrastructures, road networks are critical, and maintaining their integrity is imperative to ensure public safety and economic ef...
Introduction
false
2501.14413
1
16,061
\label{related-works} Early attempts at pavement crack detection primarily relied on low-level image properties such as gradient, brightness, shape, and texture, as well as pixel intensity variance, edge orientation, and local binary patterns. Classic edge detection algorithms such as Sobel and Canny \cite{NHATDUC2018...
Related Works
false
2501.14413
2
16,062
\subsection{Problem Structure and Overview} Detecting small and tiny cracks in pavement images is particularly challenging due to their subtle features, irregular shapes, and the presence of noise such as shadows and debris. Existing deep learning models often struggle with capturing these fine-grained details, unders...
Method
false
2501.14413
3
16,063
To validate the effectiveness of the proposed \textit{Context-CrackNet}, we conduct comprehensive experiments designed to evaluate its performance on pavement distress segmentation tasks. This section details the datasets used, the implementation specifics of \textit{Context-CrackNet}, and the experimental setup. Furt...
Experiments
false
2501.14413
4
16,064
This section presents the results of the proposed \textit{Context-CrackNet} across multiple datasets. Both qualitative and quantitative evaluations are discussed, highlighting the model's performance in comparison with existing state-of-the-art methods. \subsection{Qualitative Analysis of Predictions} In this section...
Results and Discussion
false
2501.14413
5
76,357
\label{sec:intro} \begin{figure}[t] \centering \scalebox{1.0}{ \begin{tabular}{c} \includegraphics[width=\linewidth]{plots/mean_corr_small.pdf} \\ \end{tabular} } \caption{Analyzing the generalization ability of some Disparity estimation methods: GWCNet~\cite{gwcnet}, CFNet~\cite...
Background
false
2505.05091
1
76,358
There exists no standardized tool for evaluating the performance of disparity estimation methods. Thus, the codebase for such a tool had to be written from scratch. In the following, we describe the benchmarking tool, \benchmark{}. Currently, it supports 4 unique architectures (new architectures to be added to \benchm...
Usage
false
2505.05091
2
76,359
\label{sec:analysis} \begin{figure*}[h] \centering \includegraphics[width=1.0\linewidth]{plots/log_all_corr_transpose.pdf} \caption{Using the FlyingThings3D dataset for disparity estimation, we perform an initial benchmarking of i.i.d. performance and generalization abilities of four popular disparity est...
Initial Evaluations using
false
2505.05091
3
95,373
\subsection*{Sequent calculus for \texttt{C}} The following sequent calculus can be found in \cite[p. 516]{OmoriWansing2020}. \vspace{0.5cm} {\Large \textbf{\texttt{G3C}}}\Large \textbf{\texttt{G3C}} \vspace{0.2cm} \textbf{Zero-premise rules:} \vspace{0.2cm} \quad \infer[\scriptstyle Rf]{\Gamma, p \Righta...
Appendix
false
2507.06854
8
223,113
We present \name, an automatic benchmark generator that constructs live, large-scale evaluation tasks from real-world open-source software by reusing natural ground truth signals—compilation results, functional test outcomes, and sanitizer checks. Instantiated on the PHP interpreter and SQLite3 engine, \namephp and \...
Conclusion
false
2505.12331
6
223,109
\label{sec:approach} \begin{figure*}[tb] \setlength{\abovecaptionskip}{5pt} \setlength{\belowcaptionskip}{0pt} \setlength{\intextsep}{0pt} \centering \includegraphics[scale=0.45]{figures/overview.pdf} \caption{Benchmark Workflow in \name} \label{fig:overview} \end{figure*} \setlength{\abovecaptionski...
Approach
false
2505.12331
2
95,368
We suppose a language $L$L consisting of propositional constants and connectives for which there are formulas specified. We will inductively define an $R$R-expression over $L$L as follows: 1. Every formula over $L$L is an $R$R-expression. 2. If $S$S is an $R$R-expression over $L$L which does not have the form $-T$-T...
The language
false
2507.06854
3
95,381
\label{appendix:detailed experimental settings} In this section, we give more details on the setup of our experiments. We evaluate the performance of our adversarial patch attack on image classification and face recognition tasks, with comparisons to state-of-the-art attack methods, such as Google Patch \citep{brown2...
Detailed Experimental Settings
false
2507.06856
8
95,379
\label{sec:further analysis} \subsection{Black-Box \& Real-World Applicability} To evaluate IAP’s generalizability beyond white-box settings, we assess both transferability and black-box performance. Adversarial patches generated on a surrogate model (e.g., ResNet-50) transfer reasonably well to unseen models like V...
Further Analyses
false
2507.06856
6
95,380
\label{sec:conclusion} We introduced IAP, a perceptibility-aware optimization framework for generating imperceptible adversarial patches. Experimental results across various computer vision tasks demonstrated our method's superiority in maintaining high stealth and strong targeted attack efficacy. Beyond the white-bo...
Conclusion
false
2507.06856
7
95,382
\label{appendix:additional experiments} \subsection{Additional Results on ImageNet} \label{appendix:additional results on Imagenet} In this section, we present additional detailed results corresponding to every victim model considered, with ``Toaster'' as the target class. The results include a comprehensive analysi...
Additional Experiments
false
2507.06856
9
95,383
\label{appendix:further analysis} \subsection{Ablation Studies} \label{appendix:ablation studies} We perform ablation studies to assess key components of IAP, including patch size, update iterations, and the regularization coefficient in the loss function (Equation \ref{eq:8}). We compare our update rule with the A...
Additional Analyses
false
2507.06856
10
352,244
Momentum \citep{nesterov1983method} is an important technique in optimization. In the context of convex and smooth optimization, Nesterov's momentum (accelerated gradient descent (AGD)) achieves the minimax optimal convergence rate \citep{nesterov2014introductory} and provably accelerates the vanilla GD method. Rece...
Introduction
false
2311.14222v1
1
76,368
\label{sec:appendix:description} Following, we describe the benchmarking tool, \benchmark{}. There exists no standardized tool for evaluating the performance of disparity estimation methods. Thus, the codebase for such a tool had to be written from scratch. In the following, we describe the benchmarking tool, \benchma...
Description of
false
2505.05091
12
76,370
\label{sec:appendix:evaluation_details} Following we provide a detailed description of the evaluation functions and their arguments provided in \benchmark{}. \subsection{Adversarial Attacks} \label{subsec:appendix:evaluation_details:adv_attack} To evaluate a model for a given dataset, on an attack, the following line...
Usage Details
false
2505.05091
14
95,384
\label{sec:introduction} Obtaining statistically significant results has long been a challenge in biometric system evaluations. Traditional approaches often involved conducting single-session tests with as many participants as possible. However, this method made it difficult to identify meaningful statistical influenc...
Introduction
false
2507.06858
1
95,385
\label{sec:related_work} \noindent In the field of biometric evaluations, traditional methodologies have predominantly relied on single-session data collection, where subjects are enrolled and authenticated within a limited time-frame. This approach, while practical, often fails to account for temporal variations in b...
Related Work
false
2507.06858
2
95,389
\noindent In this work, we have presented a comprehensive controlled longitudinal evaluation of facial biometrics, spanning nearly three years and over 2476 high-quality frontal images. The controlled \gls{bez}bez setup allowed us to evaluate temporal effects in frontal face images. The resulting dataset exhibits exce...
Conclusion
false
2507.06858
6
95,388
\noindent The results of this longitudinal analysis confirm that facial recognition performance is affected by temporal variation. Across all four algorithms tested, a general decline in similarity was observed over time. While the trend was present in all algorithms, the magnitude of degradation varied significantly....
Discussion
false
2507.06858
5
76,384
\begin{table*}[htbp] \centering \caption{Notations.}\label{tab:notations} \begin{tabular}{@{}ll@{}} \toprule Symbol & Description \\ \midrule $n$ & Number of clients \\ $K$ & Number of local SGD \\ $m$ & Number of selected clients\\ $m'$ & Maximum permissible age\\ $L$ & Lip...
Assumptions and Definitions
false
2505.05099
6
76,385
In this section, we evaluate the model's load distribution, convergence, and accuracy in scenarios where only a subset of clients $(15\%)$(15\%) participate in each communication round. We compare four client selection methods: random selection, probabilistic selection based on dataset size, our proposed decentralized...
Numerical Results
false
2505.05099
7
76,386
This paper presents a novel Age of Information (AoI)-based client selection mechanism designed to directly address load imbalance and slow convergence. By employing a decentralized Markov scheduling policy, the proposed approach ensures balanced client participation and minimizes the variance in selection intervals. T...
Conclusions
false
2505.05099
8
76,409
\label{sec:intro} \par \Ac{mg}mg methods are powerful tools for enhancing the computational efficiency of numerical methods. They work by solving the problem at multiple levels of resolution, smoothing out high-frequency errors on coarser grids and then refining the solution on finer grids \cite{Briggs2000,trottenberg...
Introduction
false
2505.05105
1
76,410
\label{sec:ghost_fem} \par Let us consider the Poisson equation with mixed boundary conditions as the model problem: \begin{equation}\label{eq:poisson_model_prob} -\boldsymbol{\Delta }u = f \text{~ in ~}\Omega, \quad u = g_D \text{~ on ~} \Gamma_D, \text{~ and } \quad \boldsymbol{n}\cdot \boldsymbol{\nabla}u = g...
The ghost finite element method
false
2505.05105
2
76,411
\label{sec:mg_method} We briefly summarize the main idea behind the \ac{mg}mg method and refer the reader to more detailed sources for an in-depth discussion (see, for example,~\cite{trottenberg2000multigrid}). We denote by $\Omega_{h}$\Omega_{h}h the set of grid nodes that are associated with $\mathcal{T}_{h}$\math...
Multigrid method
false
2505.05105
3
95,390
\noindent While this study provides valuable insights into the temporal stability and variability of facial biometrics over a three-year period, several avenues for future research remain open. First, as the study at the \gls{bez}bez continues, the increasing volume of data per subject will enable more precise and s...
Outlook
false
2507.06858
7
95,392
\subsection{Using AI/ML in ADHD support: Trends, diagnosis vs assistive models} There are many applications to support ADHD which have been built using AI and ML technologies and algorithms. They have been focussed on diagnosis, classification, symptom prediction. For example, studies such as those by Chen et al. (2...
Literature Review
false
2507.06864
2
95,393
The development of this ADHD-focused productivity assistant is based on a multi-dimensional approach that aligns with the central hypothesis: that attention challenges faced by ADHD-affected professionals in digital work environments can be mitigated through adaptive, human-centered and systems-aware interventions. To...
Methodology
false
2507.06864
3
76,423
Real-world embodied tasks often involve multiple agents or human-AI collaboration and competition, where dynamic interactions among agents in a shared environment give rise to emergent group-level behaviors that individuals cannot achieve alone. As a result, directly transferring methods designed for single-agent se...
Multi-agent Embodied AI
false
2505.05108
4
76,424
\label{sec:disscussions} Despite of the rapid progress of embodied AI and the primary development of multi-agent embodied AI, it faces several challenges and presents exciting future directions. \paragraph{Theory for complex Embodied AI interaction.}Theory for complex Embodied AI interaction. Building upon the t...
The Challenges and Future Works
false
2505.05108
5
76,426
Large Language Models (LLMs)~\citep{openai2024gpt4technicalreport,grattafiori2024llama3herdmodels,qwen2025qwen25technicalreport} exhibit impressive abilities in various domains such as text generation~\citep{openai2024gpt4technicalreport,grattafiori2024llama3herdmodels,xu2025personalized}, instruction following~\citep...
Introduction
false
2505.05111
1
95,394
Based on the methodology of systems-oriented design and participatory insights discussed in the methodology, this section presents the core architecture of the proposed assistant. The framework integrates lightweight on-device sensing, voice-based interaction and adaptive, context-sensitive feedback mechanisms designe...
Proposed Framework: A privacy-first and neuro-inclusive productivity assistant
false
2507.06864
4
95,395
To anchor the assistant to the everyday experiences of ADHD-affected professionals, a survey was conducted with 25 ADHD-affected digital professionals, primarily from the software engineering and IT services industry. Participants ranged from 21 to 54 years and included both professionally diagnosed and self-diagnosed...
Survey insights and user informed design decisions
false
2507.06864
5
95,396
As digital work environments continue to evolve, so too must the tools that support cognitive regulation, especially for neurodivergent professionals. This section reflects on the broader implications of our proposed assistant, exploring how it challenges conventional paradigms of productivity, how it may reshape coll...
Discussion and Implications
false
2507.06864
6
95,397
\subsection{Limitations} While this paper introduces a novel, privacy-first cognitive assistant for ADHD-affected professionals, it is not without limitations. First, the system is currently conceptual and in a very early stage of prototyping, with no deployment data. While initial user preferences were captured via ...
Limitations and Future work
false
2507.06864
7
95,398
As digital work environments become increasingly fragmented and cognitively demanding, professionals affected by ADHD continue to navigate persistent challenges—task switching, time blindness, emotional dysregulation and executive dysfunction—often without meaningful support. Our survey findings echo what lived experi...
Conclusion
false
2507.06864
8
95,407
\label{sec:prelim} \subsection{Preliminaries} Let $\mathcal{H}_A$\mathcal{H}_A be a Hilbert space of dimension $d_A$d_A and $\tau_A\in \mathcal{D}(\mathcal{H}_A)$\tau_A\in \mathcal{D}(\mathcal{H}_A) the density matrix of a quantum state, $\mathcal{D}(\mathcal{H}_A)$\mathcal{D}(\mathcal{H}_A) being the set of all densit...
Introduction
false
2507.06868
1
95,408
Let us define the following quantity: \begin{align}\label{eq:ranges} \Etrccclb= \begin{cases} \EexRR+R \ \ \text{ for } R \leq R^*\\ \Er(R,Q) \ \ \text{ for } R > R^* \end{cases} \end{align}\begin{align}\label{eq:ranges} \Etrccclb= \begin{cases} \EexRR+R \ \ \text{ for } R \leq R^*\\ \Er(R,Q) \ \ \text{ for } R > R^* ...
Main Result
false
2507.06868
2
95,409
In this section we evaluate numerically the bound derived in Theorem \ref{theo:1} and its dependency on relevant channel parameters. We consider a binary \ac{CQ}CQ channel with a symmetric input distribution, that is, $Q(1)=Q(2)=1/2$Q(1)=Q(2)=1/2. Let $\Pur=\rm{Tr}\{(\sigma^x)^2\}$\Pur=\rm{Tr}Tr\{(\sigma^x)^2\} be the...
Numerical Results
false
2507.06868
3
95,410
\subsection{Proof of Lemma \ref{lemma:ex}} We start by bounding from above the tilted error probability: \begin{align}\label{eq:ex_chain} \Pecn^\frac{1}{\Grho} &= \left(\frac{1}{M_n}\sum_{m=1}^{M_n}\PecnmallMdet \right)^\frac{1}{\Grho}\\ &\leq \frac{1}{M_n^\frac{1}{\Grho}}\sum_{m=1}^{M_n}\PecnmallMdet ^\frac{1}{\Grh...
Appendix
false
2507.06868
4
95,411
Port-Hamiltonian (pH) systems have been developed as a theoretical framework for modeling, numerical simulation, and control of macroscopic multiphysics systems. PH systems are a generalization of classical Hamiltonian systems to open systems, embedding additional pairs of port variables. These are interface variable...
Introduction
false
2507.06869
1
854
In this paper, we only consider simple and finite graphs. For a graph $G=\left( V,E\right)$G=\left( V,E\right) and a vertex $v\in V$v\in V, the degree of $v$v is denoted by $d_G\left( v\right)$d_G\left( v\right). The maximum and minimum degrees of $G$G are denoted by $\Delta(G)$\Delta(G) and $\delta(G)$\delta(G) resp...
Introduction
false
2508.00483
1
855
Let $G=(V,E)$G=(V,E) be a graph and $r\geq 2$r\geq 2 be an integer. Suppose that $v\in V$v\in V and $e\in E$e\in E. Let $N_G(v)$N_G(v) be the neighborhood of $v$v in $G$G. We denote by $\mathcal{N}_{v}(H,G)$\mathcal{N}_{v}v(H,G) (respectively, $\mathcal{N}_{e}(H,G)$\mathcal{N}_{e}e(H,G)) the number of $H$H-copies tha...
Preliminaries
false
2508.00483
2
856
In this section, we prove Theorems~\ref{Many_Kr_indicates_subgraph_with_many_Kr},~\ref{Upperbound_of_Kr_in_some_F_free} and~\ref{Lowerbound_of_Kr_in_some_F_free}. In the proof Theorem~\ref{Many_Kr_indicates_subgraph_with_many_Kr}, we first simultaneously remove edges that are not contained in many $K_r$K_r-copies....
Proof of the main theorems
false
2508.00483
3
76,427
\paragraph{SAEs.}SAEs. \begin{figure*}[t] \setlength{\abovecaptionskip}{-0.10cm} \setlength{\belowcaptionskip}{-0.30cm} \centering \includegraphics[width=\textwidth]{figs/histogram_gemma-2-2b_layer_20.pdf} \caption{The values of $\nu$, as referenced in Eq.~\ref{eq:ablation}, where a larger $\nu$ indicates stro...
Preliminary
false
2505.05111
2
76,428
\label{sec:lan_specific_feature} \subsection{Finding Language-Specific Features} To find language-specific features, we conduct a preliminary experiment by prompting Flores-10 into the LLMs and analyzing the residual stream using SAEs. We find that the mean activation of some features is particularly high for a cer...
Language-Specific Features
false
2505.05111
3
76,429
In earlier sections, we only evaluate language-specific features on monolingual texts. This raises a natural question: are these language-specific features solely related to language-specific tokens? To explore this, we focus on a phenomenon called ``code-switching.''\footnote{Code-switching refers to the practice of ...
Language-Specific Features Extend Beyond Language-Specific Tokens
false
2505.05111
4
76,430
\label{sec:ablate} \begin{figure*}[t] \setlength{\abovecaptionskip}{-0.10cm} \setlength{\belowcaptionskip}{-0.50cm} \centering \includegraphics[width=\textwidth]{figs/line_chart_ce_loss_gemma-2-2b_all_main.pdf} \caption{The change in CE loss for three languages after ablating French Features. Simultaneously ab...
Ablating Language-Specific Features Leads to Language-Specific Changes
false
2505.05111
5
95,416
In this work, we have proposed a structure-preserving spatial discretization method for pH systems with implicit constitutive relations, formulated through Stokes-Lagrange structures. Building upon the Partitioned Finite Element Method (PFEM), which has been successfully employed for classical pH systems, we have ext...
Conclusion
false
2507.06869
6
95,417
This work was supported by the IMPACTS project funded by the French National Research Agency (project ANR-21-CE48-0018), \url{https://impacts.ens2m.fr/}.
Acknowledgements
false
2507.06869
7
95,419
\label{apx:vector-identities} \begin{lemma} Given $\bm{\phi} \in H^\curl(\Omega)$ and $\psi \in H^1(\Omega)$, we have: \begin{equation} \label{ipp:gradperp-curl} \int_\Omega \bm{\phi} \cdot \grad^\perp(\psi) \, \d x = \int_\Omega \curldeuxD(\bm{\phi}) \, \psi \, \d x + \int_{\partial \Omega} \psi \, \, \bm{\phi} \...
Useful identities
false
2507.06869
9
95,412
\label{sec:sec2} The geometric definition of pH systems introduces two structures: the Dirac structure that describes the energy routing of the system, and the Lagrange structure that describes the Hamiltonian of the system. In the linear time-invariant finite-dimensional case, these structures correspond to algebrai...
Dirac and Lagrange structures
false
2507.06869
2
95,420
\subsection{Condition number} \label{apx:nanorod-condition-number} Figure \ref{fig:nanorod-condition-number} presents the condition number of $\bm{M} + \ell^2 \, \bm{K} + \ell \, \bm{B}\bm{B}^\top$\bm{M}M + \ell^2 \, \bm{K}K + \ell \, \bm{B}B\bm{B}B^\top as a function of the parameter $\ell$\ell, in particular for ...
Nanorod equation
false
2507.06869
10
95,421
\subsection{Time discretization} Similarly to \cite{de2019inclusion}, we will consider a staggered approach where $\overline{\psi}$\overline{\psi}, $u_D^3$u_D^3 and $\tilde u_D$\tilde u_D are evaluated at half time steps $t_{k+1/2}$t_{k+1/2}k+1/2 and where $\overline{\omega}$\overline{\omega} and $u_D^5$u_D^5 are ...
Incompressible Navier-Stokes equation
false
2507.06869
11
95,413
\label{sec:continuous-models} \subsection{Nanorod equation} \label{sec:nanorod} Given a domain $\Omega = (a,b)$\Omega = (a,b), let us consider the following 1D equation that describes the longitudinal motion of a nanorod \cite{maschke2023linear}: \begin{equation} \label{eqn:nanorod-equation-no-const} \partial_t\b...
Continuous models
false
2507.06869
3
95,414
\label{sec:PFEM} This section aims at presenting structure-preserving spatial discretizations of the three previous models. In particular, this procedure yields discrete Dirac and Lagrange structures \cite{gernandt2022equivalence}, corresponding to their infinite-dimensional Stokes-Dirac and Stokes-Lagrange counterpa...
Spatial Discretization
false
2507.06869
4
95,415
\label{sec:numerics} Hereafter, numerical results were obtained using the Python interface of the finite element library GetFEM \cite{renard2020getfem} and \href{https://gmsh.info/}https://gmsh.info/{Gmsh}Gmsh for mesh generation. Visualization was carried out using \href{https://www.paraview.org/}https://www.paraview...
Numerical results
false
2507.06869
5
95,422
\label{sec:intro} Personalized recommender systems aim to suggest potentially attractive items based on users' profiles and behavior histories, playing a crucial role in alleviating information overload and enhancing user experience. As platforms such as Amazon and Taobao expand, the number of recommendation domains ...
Introduction
false
2507.06877
1
95,423
\subsection{Multi-Domain Recommendation} Multi-domain recommendation (MDR) uses domain knowledge transfer to enhance performance across all included domains. Inspired by multi-task learning \cite{ma2018mmoe, tang2020ple}, many methods, such as STAR \cite{sheng2021star}, AESM\textsuperscript{2}2 \cite{zou2022AESM2}, ...
Related Work
false
2507.06877
2
95,424
\subsection{Problem Definition} We focus on the domain clustering phase in MDR, a critical step before model training. Given \(D\)D domains \(\mathcal{D} = \{d_1, d_2, \ldots, d_D\}\)\mathcal{D} = \{d_1, d_2, \ldots, d_D\}, we aim to partition them into \(K\)K clusters. Unlike conventional domain clustering m...
Preliminary
false
2507.06877
3
857
Applying the theorems in the previous section, we can obtain upper and lower bounds of $\mathrm{mex}(m,K_r,F)$\mathrm{mex}(m,K_r,F) for various families of graphs. \subsection{$K_{s,t}$-free graphs} In this subsection, we prove Corollary \ref{UpperBound_K_r_in_K_st-free_Fixed_Edge}. The following lemma establishes a...
Applications
false
2508.00483
4
76,438
\label{Appendix:nu} \begin{figure*}[t] % \setlength{\abovecaptionskip}{0.10cm} % \setlength{\belowcaptionskip}{-0.30cm} \centering \includegraphics[width=\textwidth]{figs/cs_data_example.pdf} \caption{Example data of our code-switching dataset.} \label{fig:cs_data} \end{figure*} \centering \includegraphics[w...
Additional Results of $$
false
2505.05111
13
95,425
\subsection{Overall Framework} (1) As shown in Figure \ref{fig:cdc}, the Causal Domain Clustering (CDC) method first learns two distinct domain affinity matrices through pre-gradient updates. The Isolated Domain Affinity Matrix $MI$MI quantifies the pairwise transfer effects across all domains without external in...
Method
false
2507.06877
4
95,426
\label{sec:exp} \subsection{Experimental Setup} \textbf{Public Dataset.} We use two widely recognized datasets: the Amazon dataset \cite{ni2019justifying} and AliCCP dataset \cite{ma2018entire}, where domains are defined based on item categories. Detailed statistics of these datasets are provided in Table \ref{tab...
Experiments
false
2507.06877
5
858
For a given graph $H$H on $h$h vertices, let $V(H) = [h]$V(H) = [h] and $s$s be a positive integer. We say that a graph $G$G is an $s$s-blow-up of $H$H if $V(G)$V(G) can be partitioned into $V_1 \cup \dots \cup V_h$V_1 \cup \dots \cup V_h, where each $V_i$V_i is an independent set, and there is an edge between $v_i \...
Concluding remarks
false
2508.00483
5
859
Yan Wang is supported by the National Key R\&D Program of China under Grant No. 2022YFA1006400 and Shanghai Municipal Education Commission (No. 2024AIYB003). Xiao-Dong Zhang is supported by the National Natural Science Foundation of China (No. 12371354) and the Science and Technology Commission of Shanghai Municipalit...
Acknowledgement
false
2508.00483
6
95,427
We conducted online A/B testing experiments over a 14-day period within our app marketplace's recommender system, \textbf{covering 64 domains and 20\% of user traffic}. This translates to \textbf{millions of users and billions of exposures daily}. Utilizing the domain clustering results $\{\mathcal{T}_k\}_{k=1}^K$\...
Online Evaluation
false
2507.06877
6
95,428
We propose a novel CDC framework to address domain clustering challenges in multi-domain recommendation systems with a large number of domains. We use two affinity matrices to dynamically capture inter-domain transfer effects in different transfer patterns. We then apply causal discovery to adaptively integrate t...
Conclusion
false
2507.06877
7
95,429
\begin{quote} \emph{“\textbf{Sapere aude}! Have courage to use your own reason!”} \\ \hfill --- Immanuel Kant, \textit{An Answer to the Question: What is Enlightenment?} (1784) \end{quote}\begin{quote} \emph{“\textbf{Sapere aude}! Have courage to use your own reason!”} \\ \hfill --- Immanuel Kant, \textit{An Answer t...
Introduction
false
2507.06878
1
95,430
\label{sec:co} The widespread adoption of AI tools, especially chatbots, in education raises concerns about how students engage with information. Rather than critically processing content, many users tend to accept AI-generated responses quickly and uncritically, preferring fast, simplified answers over independent re...
Cognitive atrophy of the learning
false
2507.06878
2
95,431
\label{sec:agency} While AI tools offer substantial support, their increasing presence in educational contexts may foster over-reliance, undermining students' ability to engage deeply and reflect critically \cite{al-zahraniUnveilingShadowsHype2024}. This dependence becomes particularly dangerous when AI-generated conte...
Loss of agency
false
2507.06878
3
95,432
\label{sec:emo} Beyond cognitive and agency effects, the integration of AI tools into education raises critical concerns about the learners' emotional well-being and identity formation. \subsection{Self-efficacy and self-esteem} Students with low academic self-efficacy or self-esteem are more likely to rely on AI to c...
Undermined emotional well-being
false
2507.06878
4
95,433
\label{sec:ethic} The integration of AI, especially chatbots, into educational environments raises profound ethical challenges, especially concerning student privacy, surveillance, academic integrity and the dynamics of power in digital learning spaces \cite{williams2024ethical, kwapiszPrivacyConcernsStudent2024}. \s...
Ethical and Academic concerns
false
2507.06878
5
95,434
\label{sec:stu} In support of a human-centered approach to the integration of AI in education, it is crucial to give voice to the students’ perspectives. \subsection{AI adoption} According to OpenAI, young adults aged 18–24 are the biggest adopters of ChatGPT in the U.S., with over one-third using the tool regularly \c...
Student perspectives
false
2507.06878
6
95,435
\label{sec:action} It is evident that students should be taught to critically assess AI educational tools to recognize its limitations \cite{zhaiEffectsOverrelianceAI2024}. The risks of AI technology are not inherent to the tools themselves but lie in how they are utilized. Therefore, effective integration requires ta...
Discussion
false
2507.06878
7
95,436
AI offers unprecedented opportunities to support, democratize and personalize education. However, it also poses risks that need to be addressed before its potential is realized. In this paper, we have structured such risks in four key areas in the learning process: cognition, emotion, ethics and agency. We advocate f...
Conclusion
false
2507.06878
8
95,447
\label{sec:conclusion} This paper presents an extension to the formal semantics of AADL RTS specified in \cite{ISOLA2022} to include time. Some of the port operations are also simplified to make the temporal semantics more concise and comprehensible to enforce the \emph{receive-execute-send} paradigm. A modal logic ...
Conclusion
false
2507.06881
10
95,454
\label{sec:BarriersMO} While \gls{mo}mo offers significant advantages for power systems, its integration into existing infrastructure does present some challenges. This section addresses these challenges, drawing insights from the successful application of \gls{mo}mo in power system optimization presented in this pap...
INTEGRATION CHALLENGES OF MANIFOLD OPTIMIZATION TECHNIQUES IN POWER SYSTEMS
false
2507.06883
7
95,455
\label{sec:VII} This paper demonstrates that \gls{mo}mo offers a robust and versatile framework for tackling complex optimization challenges, particularly the non-convex power flow problems inherent in \gls{epds}epds and \gls{epts}epts. \gls{mo}mo has shown significant advantages over conventional methods by leveragi...
CONCLUSION
false
2507.06883
8
95,448
\IEEEPARstart{T}T{he}he core strength of \gls{mo}mo methods lies in their superior ability to tackle complex optimization problems, outperforming conventional methods in efficiency and effectiveness, particularly for the intricate non-convex optimization challenges inherent in \gls{epds}epds and \gls{epts}epts with ...
INTRODUCTION
false
2507.06883
1
95,449
\label{sec:II} In this section, we start by highlighting different alternatives to the \gls{mo}mo technique. In the subsequent subsection, we explain the \gls{mo}mo framework. Finally, we catalog a list of manifolds found in many different real-world problems. \subsection{ALTERNATIVES TO MANIFOLD OPTIMIZATION TECHNI...
MANIFOLD FUNDAMENTALS
false
2507.06883
2
95,450
\label{sec:III} While different optimization tasks in power systems operate under established physical laws and network principles, specific problems encountered within these scenarios will invariably exhibit unique characteristics, involve distinct operational data, and possess varying properties. To effectively mod...
MANIFOLD LEARNING METHODS AND STRUCTURES
false
2507.06883
3
95,451
\label{sec:III} In the following, we illustrate the \gls{mo}mo methodology by detailing the steps involved in addressing a real-world problem in power systems. Specifically, we will optimize the voltage phase angles in a distributed grid using a \gls{mo}mo technique. This approach serves as an example, but the methodo...
METHODOLOGY FOR FORMULATING AND SOLVING OPTIMIZATION PROBLEMS WITH MANIFOLDS
false
2507.06883
4