Thesis Tide - Find relevant and interesting papers

A Theoretical Analysis of Soft-Label vs Hard-Label Training in Neural Networks

Knowledge distillation, where a small student model learns from a pre-trained large teacher model, has achieved substantial empirical success since the seminal work of \citep{hinton2015distilling}. De...

Useful Fields:

This article presents a thorough theoretical analysis that elucidates the advantages of using soft-labels in neural network training compared to hard-labels, which is crucial for understanding model efficiency and performance. It combines empirical results with robust theoretical contributions, addressing a key aspect of model training that could influence future research and applications in machine learning.

DISHONEST: Dissecting misInformation Spread using Homogeneous sOcial NEtworks and Semantic Topic classification

The emergence of the COVID-19 pandemic resulted in a significant rise in the spread of misinformation on online platforms such as Twitter. Oftentimes this growth is blamed on the idea of the "ech...

Useful Fields:

The study presents a novel approach to examining the complex dynamics of misinformation spread on social media, particularly during the COVID-19 pandemic. Its methodological rigor, utilizing both social network analysis and semantic topic classification, provides robust insights into combative information interactions. The link between user behavior and content diversity is a significant contribution that could inform strategies to combat misinformation.

Engineering micromotion in Floquet prethermalization via space-time symmetries

We present a systematic framework for Floquet prethermalization under strong resonant driving, emphasizing the pivotal role of dynamical space-time symmetries. Our approach demonstrates how dynamical ...

Useful Fields:

The article introduces a novel framework for understanding Floquet prethermalization, emphasizing the role of dynamical space-time symmetries, which is a significant advancement in the field of out-of-equilibrium quantum systems. The methodological rigor is strong, with clear protocols for implementation and observation. Its potential for inspiring future research is high due to the introduction of new techniques for detecting symmetries and the implications for understanding dynamic behavior in quantum systems.

Characterizing maximally many-body entangled fermionic states by using $M$ -body density matrix

Fermionic Hamiltonians play a critical role in quantum chemistry, one of the most promising use cases for near-term quantum computers. However, since encoding nonlocal fermionic statistics using conve...

Useful Fields:

The article presents a significant advancement in understanding many-body entanglement in fermionic systems, which is crucial for quantum computing applications. Its novel application of hypergraph structures and rigorous analysis of $M$-body reduced density matrices contributes to the depth of theoretical physics and quantum chemistry. The connection to random matrix theory further enhances its interdisciplinary relevance. The methodological rigor, particularly the semi-analytical demonstrations, adds robustness to the findings and their applicability in practical scenarios within quantum computation and chemistry.

Limits on dark matter, ultralight scalars, and cosmic neutrinos with gyroscope spin and precision clocks

Dark matter (DM) within the solar system induces deviations in the geodetic drift of gyroscope spin due to its gravitational interaction. Assuming a constant DM density as a minimal scenario, we const...

Useful Fields:

This article provides significant new constraints on dark matter and ultralight scalar fields—a novel approach that leverages existing technology in precision timing and gyroscopes. The methodology is rigorous and opens up new avenues for both theoretical exploration and experimental verification, potentially influencing the search for dark matter. The implications for astrophysics and cosmology are profound, especially in high-precision scenarios where traditional detection methods face limitations.

Omnidirectional shuttling to avoid valley excitations in Si/SiGe quantum wells

Conveyor-mode shuttling is a key approach for implementing intermediate-range coupling between electron-spin qubits in quantum dots. Initial shuttling results are encouraging; however, long shuttling ...

Useful Fields:

This article presents novel theoretical approaches to address a key limitation in quantum dot technologies by reducing valley excitations in Si/SiGe quantum wells, a significant advancement in quantum computing research. The exploration of omnidirectional shuttling schemes and their potential integration into qubit architectures is methodologically rigorous, with simulations supporting their efficacy. Its results could inspire further experimental validations and innovations in qubit design and couplings.

FreeSplatter: Pose-free Gaussian Splatting for Sparse-view 3D Reconstruction

Existing sparse-view reconstruction models heavily rely on accurate known camera poses. However, deriving camera extrinsics and intrinsics from sparse-view images presents significant challenges. In t...

Useful Fields:

This article presents a novel approach to 3D reconstruction, addressing significant challenges associated with sparse-view images and camera parameter estimation. The use of a streamlined transformer architecture is innovative and suggests substantial improvements in both efficiency and output quality. The high quality of reconstruction and the versatility of the framework across various application domains significantly enhance its impact.

DiverseAgentEntropy: Quantifying Black-Box LLM Uncertainty through Diverse Perspectives and Multi-Agent Interaction

Quantifying the uncertainty in the factual parametric knowledge of Large Language Models (LLMs), especially in a black-box setting, poses a significant challenge. Existing methods, which gauge a model...

Useful Fields:

The article addresses a pressing challenge in the field of artificial intelligence concerning the reliability and uncertainty of Large Language Models (LLMs). Its novel approach, DiverseAgentEntropy, utilizes multi-agent interaction to better assess model uncertainty, which presents a significant advancement over existing self-consistency methods. The methodological rigor and the introduction of an abstention policy to enhance reliability can have practical implications for deploying LLMs in real-world scenarios. The findings regarding hallucination detection also contribute to the ongoing discourse on model evaluation. Overall, the article provides substantial novelty and potential for influencing future research directions in LLM uncertainty evaluation.

Asymmetric Temperature Variations In Protoplanetary disks: I. Linear Theory and Corotating Spirals

Protoplanetary disks can exhibit asymmetric temperature variations due to phenomena such as shadows cast by the inner disk or localized heating by young planets. We have performed both linear analyses...

Useful Fields:

This article presents a novel approach to understanding temperature variations in protoplanetary disks, linking them to spiral structures in a robust theoretical framework. The combination of linear analyses and hydrodynamical simulations provides a rigorous methodological foundation. The findings about density and velocity perturbations have significant implications for disk accretion processes, which is a central issue in astrophysics. The exploration of boundary conditions and their effects enhances the article's depth, presenting potential observational consequences, thereby making it impactful for future studies in the field.

Arbitrary Spectral Edge of Regular Graphs

We prove that for each $d\geq 3$ and $k\geq 2$ , the set of limit points of the first $k$ eigenvalues of sequences of $d$ -regular graphs is \[ \{(μ_1,\dots,μ_k): d=μ...

Useful Fields:

This article addresses a significant conjecture in spectral graph theory, providing a rigorous proof that advances understanding of eigenvalues in regular graphs. The methodology of utilizing an infinite random graph is innovative, potentially paving the way for new models and theories in the field. The established link to existing work by well-known figures such as Alon and Wei and Huang and Yau indicates both relevance and a strong academic foundation, which enhances the paper's impact.

JuStRank: Benchmarking LLM Judges for System Ranking

Given the rapid progress of generative AI, there is a pressing need to systematically compare and choose between the numerous models and configurations available. The scale and versatility of such eva...

Useful Fields:

The article introduces a novel framework for evaluating Large Language Model (LLM) judges with a focus on their performance in system ranking rather than just response assessment. This represents a significant advancement in the field of generative AI, emphasizing the importance of judge behavior in producing meaningful comparisons across systems. The methodological rigor of conducting large-scale studies adds robustness to the findings, making it a pivotal resource for both researchers and practitioners.

A glitch in gravity: cosmic Lorentz-violation from fiery Big Bang to glacial heat death

One regime where we might see departures from general relativity is at the largest accessible scales, with a natural choice in cosmology being the cosmological horizon (or Hubble) scale. We investigat...

Useful Fields:

The article presents a novel approach to addressing significant unresolved issues in cosmology related to gravitational dynamics, specifically by proposing a single-parameter extension of the standard model. The findings are timely, as they address current tensions in cosmological measurements, such as the Hubble tension, and are supported by robust observational data. This theoretical framework has the potential to advance understanding in fundamental physics, making it highly impactful for future research.

Temporal Triadic Closure: Finding Dense Structures in Social Networks That Evolve

A graph G is c-closed if every two vertices with at least c common neighbors are adjacent to each other. Introduced by Fox, Roughgarden, Seshadhri, Wei and Wein [ICALP 2018, SICOMP 2020], this definit...

Useful Fields:

The article presents a novel approach to understanding the dynamics of temporal networks through the introduction of temporal c-closure. The definition proposed helps bridge existing static models with real-world temporal data, which is a progressive step in social network research. The methodology for enumerating maximal cliques within these temporal structures adds significant value, suggesting practical applications in analyzing network evolution. The experimental validation with real-world datasets enhances its credibility, though further exploration in diverse temporal contexts could strengthen its impact.

How Many Bursts Does it Take to Form a Core at the Center of a Galaxy?

We present a novel method for systematically assessing the impact of central potential fluctuations associated with bursty outflows on the structure of dark matter halos for dwarf and ultra-faint gala...

Useful Fields:

The article introduces a novel method using dark matter simulations to explore core formation in dwarf galaxies, addressing a critical issue in astrophysics related to dark matter halo structures. The methodological rigor, with controlled simulations and direct applicability to observed phenomena, enhances its significance. Its findings have broad implications for understanding galaxy evolution and the nature of dark matter, making it highly relevant for ongoing and future research in astrophysics.

Obfuscated Activations Bypass LLM Latent-Space Defenses

Recent latent-space monitoring techniques have shown promise as defenses against LLM attacks. These defenses act as scanners that seek to detect harmful activations before they lead to undesirable act...

Useful Fields:

This paper identifies a significant vulnerability in existing latent-space defenses against LLM (Large Language Model) attacks. It presents novel findings on obfuscated activations that have substantial implications for the security of LLM implementations. The rigorous investigation into the effectiveness of current defenses against malicious manipulation positions the study as both relevant and impactful for ongoing research in model security.

Improving the Reliability of Cable Broadband Networks via Proactive Network Maintenance

Cable broadband networks are one of the few "last-mile" broadband technologies widely available in the U.S. Unfortunately, they have poor reliability after decades of deployment. The cable i...

Useful Fields:

The article presents a novel system, CableMon, that significantly enhances the reliability of cable broadband networks by leveraging machine learning techniques on Proactive Network Maintenance (PNM) data. The research addresses a critical gap in the existing tools that have high false-positive rates, showcasing methodological rigor through a strong evaluation process with real-world data. This advancement is likely to inspire further research in network maintenance and machine learning applications in telecommunications.

Does Representation Matter? Exploring Intermediate Layers in Large Language Models

Understanding what defines a good representation in large language models (LLMs) is fundamental to both theoretical understanding and practical applications. In this paper, we investigate the quality ...

Useful Fields:

This article scores highly due to its rigorous exploration of intermediate representations in LLMs, an area that has received limited attention relative to final layer outputs. The methodology employs various innovative metrics, suggesting a strong empirical foundation. By revealing significant architectural variations and the influence of input parameters on representation quality, the findings could drive future research in model optimization and training strategies. Overall, this work is both novel and applicable across several domains.

Identification of structures driving trailing-edge noise. Part II -- Numerical investigation

The aim of the present work is to investigate the mechanisms of broadband trailing-edge noise generation to improve prediction tools and control strategies. We focus on a NACA 0012 airfoil at 3 degree...

Useful Fields:

This article offers a significant advancement in understanding trailing-edge noise through rigorous numerical simulations and comparison with experimental data. The use of high-fidelity LES techniques and innovative analysis methods such as SPOD highlights its methodological rigor and depth. The novelty lies in confirming the nature of turbulent structures contributing to noise generation, which can impact future aerodynamic designs and noise reduction strategies.

Two-dimensional orbital-obstructed insulators with higher-order band topology

Obstructed atomic phases, with their realizations in systems of diverse dimensionality, have recently arisen as one of the topological states with greatest potential to show higher-order phenomena. In...

Useful Fields:

This article presents a novel exploration of orbital-obstructed insulators, focusing on higher-order band topology. The use of a tight-binding model combined with first-principles calculations demonstrates methodological rigor. The findings could significantly advance the understanding of topological phases in two-dimensional materials, with broad implications for both theoretical and applied physics.

Foundational Large Language Models for Materials Research

Materials discovery and development are critical for addressing global challenges. Yet, the exponential growth in materials science literature comprising vast amounts of textual data has created signi...

Useful Fields:

This article presents a highly relevant advancement in materials research through the development of LLaMat, a specialized Large Language Model tailored for processing materials science literature. Its focus on domain-specific adaptation and the demonstrated success in crystal structure prediction represent significant contributions to the field. The implications for other domains also enhance its relevance. Methodological rigor is evident in the systematic evaluation of models, making its findings robust and applicable.