Thesis Tide - Find relevant and interesting papers

Rumor Detection on Social Media with Temporal Propagation Structure Optimization

Traditional methods for detecting rumors on social media primarily focus on analyzing textual content, often struggling to capture the complexity of online interactions. Recent research has shifted to...

Useful Fields:

The article addresses a significant gap in existing rumor detection frameworks by integrating temporal information and optimizing the propagation structure. Its emphasis on the temporal dimension and noise reduction adds both novelty and robustness to the methodology, potentially leading to improved real-world applications. Moreover, the usage of graph neural networks indicates a strong methodological rigor. The empirical validation on well-established datasets enhances the credibility of the findings.

Adaptive Prompting for Continual Relation Extraction: A Within-Task Variance Perspective

To address catastrophic forgetting in Continual Relation Extraction (CRE), many current approaches rely on memory buffers to rehearse previously learned knowledge while acquiring new tasks. Recently, ...

Useful Fields:

This article presents a novel approach to a significant problem in the field of Continual Relation Extraction, specifically addressing critical limitations of existing methods. The proposed method offers innovation by integrating prompt pools and a generative model, promoting both task-specific and cross-task learning while mitigating forgetting—critical for advancing this area of research. The experimental validation suggests methodological rigor and potential practical applicability, thus making it valuable for further studies and applications in related domains.

LCFO: Long Context and Long Form Output Dataset and Benchmarking

This paper presents the Long Context and Form Output (LCFO) benchmark, a novel evaluation framework for assessing gradual summarization and summary expansion capabilities across diverse domains. LCFO ...

Useful Fields:

The LCFO benchmark introduces a novel framework that enhances the evaluation of summarization and summary expansion, which are critical areas in natural language processing. Its comprehensive dataset and multi-faceted evaluation metrics improve the robustness of existing models and promote rigorous testing against human quality standards. The focus on long documents adds significant value, as conventional benchmarks often prioritize shorter texts, marking a notable advancement in the field.

TouchTTS: An Embarrassingly Simple TTS Framework that Everyone Can Touch

It is well known that LLM-based systems are data-hungry. Recent LLM-based TTS works typically employ complex data processing pipelines to obtain high-quality training data. These sophisticated pipelin...

Useful Fields:

The article introduces a novel approach to TTS (Text-to-Speech) systems by simplifying the data processing pipeline and addressing cost and quality challenges associated with LLMs. Its major strengths are the innovative use of the S3Tokenizer for enhanced data retention and quality, as well as a unified model architecture that can streamline deployment. These improvements are likely to encourage further research into more efficient TTS systems and could inspire interdisciplinary applications in both natural language processing and audio processing domains.

Efficient estimation of error bounds for quantum multiparametric imaging with constraints

Advanced super-resolution imaging techniques require specific approaches for accurate and consistent estimation of the achievable spatial resolution. Fisher information supplied to Cramer-Rao bound (C...

Useful Fields:

This article offers a novel approach to overcoming a significant limitation of the conventional Cramer-Rao bound in the context of quantum imaging, specifically when dealing with constraints. The development of an algorithm for constructing a modified Fisher information matrix is a substantive contribution that improves theoretical foundations and practical applications in the field. Its rigorous demonstration across several model problems, alongside the applicability to real-world experiments, strengthens its relevance and potential impact.

Mathematical description of continuous time and space replicator-mutator equations for quadratic fitness landscapes

The replicator-mutator equation is a model for populations of individuals carrying different traits, with a fitness function mediating their ability to replicate, and a stochastic model for mutation. ...

Useful Fields:

The article presents a novel analytical approach to the replicator-mutator equations, which enhances understanding of evolutionary dynamics in continuous trait spaces. The derivation of solutions for complex fitness landscapes without reliance on simulations offers significant methodological advancements. The exploration of various evolutionary phenomena indicates high applicability and potential for influencing future research directions in evolutionary biology and mathematical modeling.

NLPineers@ NLU of Devanagari Script Languages 2025: Hate Speech Detection using Ensembling of BERT-based models

This paper explores hate speech detection in Devanagari-scripted languages, focusing on Hindi and Nepali, for Subtask B of the CHIPSAL@COLING 2025 Shared Task. Using a range of transformer-based model...

Useful Fields:

The paper showcases a novel approach to a critical social issue—hate speech detection—in underrepresented languages, demonstrating methodological rigor through the use of various BERT-based transformer models. The focus on Devanagari-scripted languages is timely and relevant, particularly given increasing internet penetration in multilingual contexts. The results indicate a competitive performance, laying groundwork for future advancements in natural language processing (NLP) applied to social media discourse.

Axion QED as a Lattice Gauge Theory and Non-Invertible Symmetry

We investigate the non-invertible symmetry associated with chiral symmetry in axion quantum electrodynamics (QED) using the modified Villain formulation. In axion QED, it is known that naive magnetic ...

Useful Fields:

This article presents a novel investigation into non-invertible symmetries in axion QED through lattice gauge theory, which contributes significantly to the understanding of chiral symmetries and gauge invariance. The methodological innovation of constructing new degrees of freedom for magnetic objects enhances its impact, suggesting strong applicability for further theoretical advancements in quantum field theory and condensed matter physics. The focus on non-standard methods of gauging also hints at new pathways for exploration.

Why Does Dropping Edges Usually Outperform Adding Edges in Graph Contrastive Learning?

Graph contrastive learning (GCL) has been widely used as an effective self-supervised learning method for graph representation learning. However, how to apply adequate and stable graph augmentation to...

Useful Fields:

The article presents a novel theoretical framework for understanding edge operations in graph contrastive learning, addressing an important gap in the literature. The introduction of the Error Passing Rate (EPR) metric is innovative and could lead to further explorations in augmentation strategies. The combination of theoretical analysis with empirical experimentation adds robustness to the findings, making it highly relevant for advancing GCL methodologies.

Parameter optimization for restarted mixed precision iterative sparse solver

We consider the problem of optimizing the parameter of a two-stage algorithm for approximate solution of a system of linear algebraic equations with a sparse $n\times n$ -matrix, i.e., with one...

Useful Fields:

This article presents a novel approach to optimizing the parameter of a two-stage algorithm in solving sparse linear systems, which is a critical area in numerical analysis and computational mathematics. The emphasis on mixed precision computation is particularly relevant in the context of enhancing performance while managing numerical accuracy, which is a key consideration in the implementation of large-scale algorithms. The methodology appears robust, utilizing regression techniques to determine optimal parameters effectively. The results could significantly influence future developments in algorithm efficiency and precision in fields that rely on sparse matrices.

From Lived Experience to Insight: Unpacking the Psychological Risks of Using AI Conversational Agents

Recent gain in popularity of AI conversational agents has led to their increased use for improving productivity and supporting well-being. While previous research has aimed to understand the risks ass...

Useful Fields:

The article presents a novel risk taxonomy that addresses psychological risks specifically connected to AI conversational agents, filling a critical gap in the existing literature. Its mixed-method approach enhances methodological rigor and the consideration of lived experiences adds depth to the findings. By proposing a multi-path vignette based framework, the study opens new avenues for understanding and mitigating psychological risks, demonstrating high applicability and potential for real-world impact. Overall, it contributes significantly to the discourse on AI safety and ethics.

Towards Foundation-model-based Multiagent System to Accelerate AI for Social Impact

AI for social impact (AI4SI) offers significant potential for addressing complex societal challenges in areas such as public health, agriculture, education, conservation, and public safety. However, e...

Useful Fields:

The article proposes a novel framework that leverages recent advancements in foundation models to streamline multi-agent systems for social impact applications, addressing a significant gap in current AI4SI research that often lacks scalability and efficiency. The focus on reducing computational costs while also considering ethical implications demonstrates a rigorous methodological approach and the potential for broad societal benefits.

$L^p$ theory for a singular Sturm-Liouville equation

In this paper we consider the following Sturm-Liouville equation \left\{ \begin{aligned} -(x^{2α}u'(x))'+u(x)&=f(x) && \text{in } (0,1],\\ u(1)&=0 \end{aligned} \right. ...

Useful Fields:

This article contributes to the field of singular Sturm-Liouville theory by exploring $L^p$ spaces, which is a critical aspect of functional analysis. The consideration of boundary conditions and the establishment of solutions offer notable advances in mathematical theory and may inspire further research into similar equations, addressing gaps in the literature regarding non-standard parameters. The methodical approach and rigorous analysis enhance its applicability, particularly to applied mathematics and theoretical physics, raising its potential importance in those areas.

Interior regularity of some weighted quasi-linear equations

In this article we study the quasi-linear equation \[ \left\{ \begin{aligned} \mathrm{div}\, \mathcal A(x,u,\nabla u)&=\mathcal B(x,u,\nabla u)&&\text{in }Ω,\\ u\in H^{1,p}_{loc}&(Ω;wd...

Useful Fields:

This article provides significant contributions to the understanding of interior regularity for weighted quasi-linear equations, which is a relatively specialized area within partial differential equations (PDEs). The establishment of regularity results for weak solutions is particularly valuable for theoretical advancements in the study of differential equations, and the connection to point-wise asymptotic estimates enhances the applicability of the results. The methods seem rigorous, and the exploration of critical exponents hints at potential broad implications for other equations. However, the highly specialized nature of the study may limit its immediate applicability to broader contexts.

Doe-1: Closed-Loop Autonomous Driving with Large World Model

End-to-end autonomous driving has received increasing attention due to its potential to learn from large amounts of data. However, most existing methods are still open-loop and suffer from weak scalab...

Useful Fields:

The article presents a novel closed-loop framework for autonomous driving, which addresses significant limitations of existing open-loop systems. Its innovative approach leveraging a large world model and multi-modal transformer shows strong potential for real-world application and scalability. The rigorous experimentation on a well-known dataset adds credibility to the claims, enhancing its impact on future developments in the field.

FreeScale: Unleashing the Resolution of Diffusion Models via Tuning-Free Scale Fusion

Visual diffusion models achieve remarkable progress, yet they are typically trained at limited resolutions due to the lack of high-resolution data and constrained computation resources, hampering thei...

Useful Fields:

The article presents a novel tuning-free approach to enhance the resolution capabilities of visual diffusion models, addressing a critical limitation in the field of image and video generation. The proposed FreeScale method effectively tackles issues related to high-frequency information and repetitive patterns in generated content, showcasing methodological rigor through extensive experimentation and validation of results. Its ability to generate 8k-resolution images distinguishes it as a significant advancement, likely inspiring further research in diffusion models and high-resolution visual generation.

Illusion3D: 3D Multiview Illusion with 2D Diffusion Priors

Automatically generating multiview illusions is a compelling challenge, where a single piece of visual content offers distinct interpretations from different viewing perspectives. Traditional methods,...

Useful Fields:

The article presents a novel approach to creating 3D multiview illusions using advanced techniques that integrate text-to-image diffusion models. This represents a significant advancement over existing methods by enhancing the artistic expressiveness and versatility of 3D illusion generation. The methodological rigor displayed in optimizing textures and geometry through differentiable rendering adds to its relevance, and the results suggest ample applicability in both artistic and technological fields.

GenEx: Generating an Explorable World

Understanding, navigating, and exploring the 3D physical real world has long been a central challenge in the development of artificial intelligence. In this work, we take a step toward this goal by in...

Useful Fields:

GenEx introduces a novel approach to AI navigation and exploration through its generative imagination system, significantly advancing the field of embodied AI. The methodology leverages a well-curated 3D world dataset and demonstrates strong results in generating consistent 3D environments, which can enhance training scenarios for AI agents. The capability of generating immersive environments from minimal input (single RGB image) is particularly innovative, suggesting potential for various practical applications. Furthermore, the integration with GPT-assisted agents promotes interdisciplinary collaboration between generative models and AI-driven navigation, making this research highly relevant for future advancements in the field.

OmniDrag: Enabling Motion Control for Omnidirectional Image-to-Video Generation

As virtual reality gains popularity, the demand for controllable creation of immersive and dynamic omnidirectional videos (ODVs) is increasing. While previous text-to-ODV generation methods achieve im...

Useful Fields:

The article presents an innovative approach to tackling the limitations of existing text-to-omnidirectional video (ODV) generation techniques, demonstrating significant advancements in motion control accuracy and quality through the proposed OmniDrag framework. The combination of pretrained models, a novel control module, and a new dataset indicates methodological rigor and real-world applicability, particularly in virtual reality contexts. The solution not only addresses existing gaps but also sets the stage for future research in ODV generation and user interaction designs.

LoRACLR: Contrastive Adaptation for Customization of Diffusion Models

Recent advances in text-to-image customization have enabled high-fidelity, context-rich generation of personalized images, allowing specific concepts to appear in a variety of scenarios. However, curr...

Useful Fields:

The article presents a highly novel approach (LoRACLR) that addresses significant limitations in customizing diffusion models for image generation, specifically the challenges related to combining multiple models without concept distortion. Its methodology demonstrates strong rigor and innovative use of contrastive objectives, which could profoundly impact future developments in this niche. The implications for personalized image generation and model efficiency lend substantial applicability to its findings.