Federated LoRA Fine-Tuning for LLMs via Collaborative Alignment

Kavli Affiliate: Feng Long| Summary: Low-rank adaptation (LoRA) has emerged as a powerful tool for parameter-efficient fine-tuning of large language models (LLMs). This paper studies LoRA under a federated learning setting, enabling collaborative fine-tuning across clients while preserving parameter efficiency. We focus on a highly heterogeneous regime in which clients share only partial structure and […]


Continue.. Federated LoRA Fine-Tuning for LLMs via Collaborative Alignment

Adaptive Test for Jump

Kavli Affiliate: Feng Long| Summary: We develop an adaptive jump test for discretely observed high-frequency semimartingales by combining the A"it-Sahalia–Jacod ratio statistic (A"it-Sahalia and Jacod, 2009) and the Lee–Mykland extreme-return statistic (Lee and Mykland, 2008) with the Cauchy combination rule. Allowing stochastic It^o drift, volatility, and leverage, we show asymptotic independence under the continuous-path null […]


Continue.. Adaptive Test for Jump

Adaptive Test for Jump

Kavli Affiliate: Feng Long| Summary: We develop an adaptive jump test for discretely observed high-frequency semimartingales by combining the A"it-Sahalia–Jacod ratio statistic (A"it-Sahalia and Jacod, 2009) and the Lee–Mykland extreme-return statistic (Lee and Mykland, 2008) with the Cauchy combination rule. Allowing stochastic It^o drift, volatility, and leverage, we show asymptotic independence under the continuous-path null […]


Continue.. Adaptive Test for Jump

WIMP-like Dark Matter Without Thermalization At Freeze-Out

Kavli Affiliate: Gordan Krnjaic | Summary:In the standard thermal relic scenario, dark matter remains in chemical equilibrium with the Standard Model radiation bath until freeze-out occurs at $T sim m_X/20$, where $m_X$ is the dark matter mass. In this familiar class of models, the observed relic density is obtained for annihilation cross sections of order […]


Continue.. WIMP-like Dark Matter Without Thermalization At Freeze-Out

TIC 295741342: A Triply-Eclipsing Triple Star System with a Giant Tertiary

Kavli Affiliate: Saul Rappaport | Summary:We present the discovery and characterization of TIC 295741342, a triply-eclipsing triple star system with a giant tertiary. The eclipsing binary consists of two similar main-sequence stars in a 4.75-day orbit. The binary is in a 412.8-day orbit with the giant tertiary. We found two degenerate solutions for the system: […]


Continue.. TIC 295741342: A Triply-Eclipsing Triple Star System with a Giant Tertiary

Bessel-Hagen currents for the Fierz-Pauli action

Kavli Affiliate: Anthony Lasenby | Summary:For electromagnetism in Minkowski spacetime, the Bessel-Hagen method gives a particularly direct Noetherian derivation of the standard gauge-invariant energy-momentum tensor. The key step is to supplement the form variation generated by an infinitesimal coordinate transformation with a compensating electromagnetic gauge transformation. In this paper we ask whether the same idea […]


Continue.. Bessel-Hagen currents for the Fierz-Pauli action

A Systematic NLTE Study of Very Metal-Poor Stars with Metallicity Down to $-4.3$ dex. II. Lithium Abundance and New Insight to the Lithium Plateau

Kavli Affiliate: Huawei Zhang | Summary:Metal-poor stars are crucially important for understanding the early Galaxy, first stars, and the Universe. In this series of papers, we present a homogeneous non-local thermodynamic equilibrium (NLTE) abundances analysis of 12 elements for 103 very/extremely metal-poor (VMP/EMP) stars with metallicity down to $-4.3$ dex. The sample was selected from […]


Continue.. A Systematic NLTE Study of Very Metal-Poor Stars with Metallicity Down to $-4.3$ dex. II. Lithium Abundance and New Insight to the Lithium Plateau

Optimal mitigation of random telegraph noise for improved photometry at high frame rates

Kavli Affiliate: Gabor Furesz | Summary:Random telegraph noise (RTN) is a major contributor to read noise in many CMOS image sensors considered for astronomical use. While scientific CMOS image sensors deliver lower read noise than traditional charge-coupled devices, mitigating RTN would widen this gap and enable more precise photometry when using the fast readout rates […]


Continue.. Optimal mitigation of random telegraph noise for improved photometry at high frame rates

The Lumina Project: CMB Optical Depth Fluctuations from Patchy Reionization

Kavli Affiliate: Mark Vogelsberger | Summary:Patchy reionization couples the ionized-bubble morphology to the underlying density field, making the CMB Thomson optical depth sensitive to both the global ionization history and anisotropic fluctuations on the sky. Using the large-volume radiation-hydrodynamical Lumina simulation, we compute $τ_rm CMB$ in two ways: (i) from global volume- and mass-weighted ionization […]


Continue.. The Lumina Project: CMB Optical Depth Fluctuations from Patchy Reionization

The Classification of the 2-generated Primitive Axial Algebras of Monster Type

Kavli Affiliate: Michael Turner | Summary:Axial algebras of Monster type are a class of commutative algebras generated by special idempotents called axes. Some motivating examples of these algebras are the Griess algebra and the Norton-Sakuma algebras, relating to the Monster simple group. A long standing open problem is to classify the 2-generated axial algebras of […]


Continue.. The Classification of the 2-generated Primitive Axial Algebras of Monster Type