Decoding Knowledge Attribution in Mixture-of-Experts: A Framework of Basic-Refinement Collaboration and Efficiency Analysis

Kavli Affiliate: Jia Liu | First 5 Authors: Junzhuo Li, Bo Wang, Xiuze Zhou, Peijie Jiang, Jia Liu | Summary: The interpretability of Mixture-of-Experts (MoE) models, especially those with heterogeneous designs, remains underexplored. Existing attribution methods for dense models fail to capture dynamic routing-expert interactions in sparse MoE architectures. To address this issue, we propose […]


Continue.. Decoding Knowledge Attribution in Mixture-of-Experts: A Framework of Basic-Refinement Collaboration and Efficiency Analysis

$δn$ formalism: A new formulation for the probability density of the curvature perturbation

Kavli Affiliate: Misao Sasaki | First 5 Authors: Diego Cruces, Shi Pi, Misao Sasaki, , | Summary: $delta N$ formalism is a useful method to calculate the curvature perturbation. Contrary to what it is typically done in the literature, we re-formulate the $delta N$ formalism by using the $e$-folding number $n$ counted forward in time. […]


Continue.. $δn$ formalism: A new formulation for the probability density of the curvature perturbation

$δn$ formalism: A new formulation for the probability density of the curvature perturbation

Kavli Affiliate: Misao Sasaki | First 5 Authors: Diego Cruces, Shi Pi, Misao Sasaki, , | Summary: $delta N$ formalism is a useful method to calculate the curvature perturbation. Contrary to what it is typically done in the literature, we re-formulate the $delta N$ formalism by using the $e$-folding number $n$ counted forward in time. […]


Continue.. $δn$ formalism: A new formulation for the probability density of the curvature perturbation

FSL-SAGE: Accelerating Federated Split Learning via Smashed Activation Gradient Estimation

Kavli Affiliate: Jia Liu | First 5 Authors: Srijith Nair, Michael Lin, Peizhong Ju, Amirreza Talebi, Elizabeth Serena Bentley | Summary: Collaborative training methods like Federated Learning (FL) and Split Learning (SL) enable distributed machine learning without sharing raw data. However, FL assumes clients can train entire models, which is infeasible for large-scale models. In […]


Continue.. FSL-SAGE: Accelerating Federated Split Learning via Smashed Activation Gradient Estimation

Exact Quantum Many-Body Scars in 2D Quantum Gauge Models

Kavli Affiliate: Masahito Yamazaki | First 5 Authors: Yuan Miao, Linhao Li, Hosho Katsura, Masahito Yamazaki, | Summary: Quantum many-body scars (QMBS) serve as important examples of ergodicity-breaking phenomena in quantum many-body systems. Despite recent extensive studies, exact QMBS are rare in dimensions higher than one. In this paper, we study a two-dimensional quantum $mathbb{Z}_2$ […]


Continue.. Exact Quantum Many-Body Scars in 2D Quantum Gauge Models

Resurgence Analysis of the Nambu-Jona-Lasinio model at large charge

Kavli Affiliate: Simeon Hellerman | First 5 Authors: Jahmall Bersini, Simeon Hellerman, Domenico Orlando, Susanne Reffert, | Summary: We study the fixed point of the three-dimensional NJL model in a double-scaling limit where both the charge $Q$ and the number of fermion flavors $N$ become large with a fixed ratio $q=Q/(2N)$. While a similar analysis […]


Continue.. Resurgence Analysis of the Nambu-Jona-Lasinio model at large charge

Nonperturbative Quantum Gravity in a Closed Lorentzian Universe

Kavli Affiliate: Yasunori Nomura | First 5 Authors: Yasunori Nomura, Tomonori Ugajin, , , | Summary: We study how meaningful physical predictions can arise in nonperturbative quantum gravity in a closed Lorentzian universe. In such settings, recent developments suggest that the quantum gravitational Hilbert space is one-dimensional and real for each $alpha$-sector, as induced by […]


Continue.. Nonperturbative Quantum Gravity in a Closed Lorentzian Universe

Nonperturbative Quantum Gravity in a Closed Lorentzian Universe

Kavli Affiliate: Yasunori Nomura | First 5 Authors: Yasunori Nomura, Tomonori Ugajin, , , | Summary: We study how meaningful physical predictions can arise in nonperturbative quantum gravity in a closed Lorentzian universe. In such settings, recent developments suggest that the quantum gravitational Hilbert space is one-dimensional and real for each $alpha$-sector, as induced by […]


Continue.. Nonperturbative Quantum Gravity in a Closed Lorentzian Universe

Finite-Time Global Optimality Convergence in Deep Neural Actor-Critic Methods for Decentralized Multi-Agent Reinforcement Learning

Kavli Affiliate: Jia Liu | First 5 Authors: Zhiyao Zhang, Myeung Suk Oh, FNU Hairi, Ziyue Luo, Alvaro Velasquez | Summary: Actor-critic methods for decentralized multi-agent reinforcement learning (MARL) facilitate collaborative optimal decision making without centralized coordination, thus enabling a wide range of applications in practice. To date, however, most theoretical convergence studies for existing […]


Continue.. Finite-Time Global Optimality Convergence in Deep Neural Actor-Critic Methods for Decentralized Multi-Agent Reinforcement Learning