arXiv subjects

Eric Davis

Publications and source records attributed to Eric Davis.

10 recordsLinked to original sources

arXiv2026

The Multipath Reliable Connection (MRC) Transport

MRC is an open, production-grade transport designed for large-scale AI/ML training over best-effort Ethernet. It extends RoCEv2 with explicit, composable primitives for per-packet multipath and sender-based congestion control, decouples packet delivery from semantic processing, adds multiple new capabilities for accelerated packet-loss recovery and adds resilience against port and path failures. This paper presents MRC and details its core capabilities and mechanisms.

cs.NI↗

arXiv2026

Resilient AI Supercomputer Networking using MRC and SRv6

Tail latency dominates the performance of synchronous pretraining jobs when running at very large scales. We describe a three-pronged approach: (1) a new RDMA-based transport protocol, MRC, sprays across many paths and actively load-balances between them, eliminating the issue of flow collisions (2) the use of multi-plane Clos topologies to get the benefits of high switch radix and redundancy, allowing training clusters well over 100K GPUs to be built as two-tier topologies while increasing physical redundancy, and (3) the use of static source-routing using SRv6 to allow MRC the freedom to bypass failures by itself. We describe our experiences running MRC and static SRv6 routing in production in OpenAI and Microsoft's largest training clusters, where it has been used to train the latest frontier models. We demonstrate how MRC allows AI training jobs to ride out many network failures that previously would have interrupted training.

cs.NI↗

arXiv2026

Gauge-Invariant Gravitational Wave Polarization in Metric f(R) Gravity with Cosmological Implications

We develop a fully gauge invariant analysis of gravitational wave polarizations in metric f(R) gravity with a particular focus on the modified Starobinsky model, whose constant curvature solution provides a natural deSitter background for both early and late time cosmology. Linearizing the field equations around this background, we derive the Klein Gordon equation for the curvature perturbation and show that the scalar propagating mode acquires a mass, highlighting how the same scalar degree of freedom governs inflationary dynamics at high curvature and the propagation of gravitational waves in the current accelerating Universe. Using the scalar vector tensor (SVT) decomposition and a decomposition of the perturbed Ricci tensor, we obtain a set of fully gauge invariant propagation equations that isolate the contributions of the scalar, vector, and tensor modes in the presence of matter. We find that the tensor sector retains the two transverse traceless polarizations of General Relativity, while the scalar sector supports a massive breathing-longitudinal mode determined by the massive scalar propagating mode. Through the geodesic deviation equation, computed both in a local Minkowski patch and in fully covariant de Sitter form, we independently recover the same polarization content and identify its tidal signatures. The resulting framework connects the extra scalar polarization to cosmological observables, providing a unified, gauge invariant link between gravitational wave phenomenology and the cosmological implications of metric f(R) gravity.

gr-qc↗

arXiv2025

Scalar-Field Wave Dynamics and Quasinormal Modes of the Teo Rotating Wormhole

We investigate scalar field perturbations of the rotating Teo wormhole. We also compute the quasinormal mode (QNM) spectrum using first order WKB approximation. After separation of variables, we obtain a Schroedinger type radial equation and a smooth barrier potential which is shaped by the localized frame-dragging effects of the wormhole throat. This barrier potential provides damped oscillatory modes for the range of spins that were examined. The QNM spectrum shows a coherent and monotonic dependence on rotation. As the spin increases, both the oscillating frequency of the scalar wave and its damping rate decrease, which indicates progressively longer lived modes in the absence of absorption due to a horizon. We have verified the correspondence in the Eikonal limit, by obtaining the radius of the photon ring, its orbital frequency, and the Lyaponov exponent. Next, we compared the Teo wormhole QNM with that of the Kerr black hole QNM to find that the Kerr QNM is dictated by absorption at the horizon and they also exhibit symmetric pro-grade retrograde mode splitting, whereas the Teo wormhole QNM shows a stronger, and spatially confined response to spin. The Teo wormhole also exhibit partial reflection at the throat and a very distinct one-sided mode splitting which rapidly saturates as the spin increases. Additionally, the rotating Teo wormhole allows an ergoregion with the possibility of frequency kinematics compatible with superradiance. Due to the absence of an event horizon or a dissipative boundary, there is no evidence of classical superradiant amplification that was seen in Kerr. The results we obtained clearly demonstrates how rotation and boundary conditions jointly shape wave propagation in horizonless compact objects. They also provide certain characteristic spectral signatures that can be used to distinguish rotating wormhole spacetimes from rotating black hole spacetimes.

gr-qc↗

arXiv2024

A Review of Stable, Traversable Wormholes in f (R) Gravity Theories

It has been proven that in standard Einstein gravity, exotic matter is required to stabilize traversable wormholes. Quantum field theory permits these violations due to the quantum coherent effects found in any quantum field. Even reasonable classical scalar fields violate the energy conditions. In the case of the Casimir effect and squeezed vacuum states, these violations have been experimentally proven. It is advantageous to investigate methods to minimize the use of exotic matter. One such area of interest is extended theories of Einstein gravity. It has been claimed that in some extended theories, stable traversable wormholes solutions can be found without the use of exotic matter. There are many extended theories of gravity, and in this review paper, we first explore modified gravity theories and then explore some wormhole solutions in such theories, including Lovelock gravity and Einstein Dilaton Gauss Bonnet gravity. For completeness, we have also reviewed other wormholes such as Casimir wormholes, dark matter halo wormholes, thin-shell wormholes, and Nonlocal Gravity wormholes, where alternative techniques are used to either avoid or reduce the amount of exotic matter that is required.

gr-qc↗

arXiv2023

WHAT, WHEN, and HOW to Ground: Designing User Persona-Aware Conversational Agents for Engaging Dialogue

This paper presents a method for building a personalized open-domain dialogue system to address the WWH (WHAT, WHEN, and HOW) problem for natural response generation in a commercial setting, where personalized dialogue responses are heavily interleaved with casual response turns. The proposed approach involves weighted dataset blending, negative persona information augmentation methods, and the design of personalized conversation datasets to address the challenges of WWH in personalized, open-domain dialogue systems. Our work effectively balances dialogue fluency and tendency to ground, while also introducing a response-type label to improve the controllability and explainability of the grounded responses. The combination of these methods leads to more fluent conversations, as evidenced by subjective human evaluations as well as objective evaluations.

cs.CL↗

arXiv2022

DLKoopman: A deep learning software package for Koopman theory

We present DLKoopman -- a software package for Koopman theory that uses deep learning to learn an encoding of a nonlinear dynamical system into a linear space, while simultaneously learning the linear dynamics. While several previous efforts have either restricted the ability to learn encodings, or been bespoke efforts designed for specific systems, DLKoopman is a generalized tool that can be applied to data-driven learning and optimization of any dynamical system. It can either be trained on data from individual states (snapshots) of a system and used to predict its unknown states, or trained on data from trajectories of a system and used to predict unknown trajectories for new initial states. DLKoopman is available on the Python Package Index (PyPI) as 'dlkoopman', and includes extensive documentation and tutorials. Additional contributions of the package include a novel metric called Average Normalized Absolute Error for evaluating performance, and a ready-to-use hyperparameter search module for improving performance.

cs.LG↗

arXiv2022

KOBEST: Korean Balanced Evaluation of Significant Tasks

A well-formulated benchmark plays a critical role in spurring advancements in the natural language processing (NLP) field, as it allows objective and precise evaluation of diverse models. As modern language models (LMs) have become more elaborate and sophisticated, more difficult benchmarks that require linguistic knowledge and reasoning have been proposed. However, most of these benchmarks only support English, and great effort is necessary to construct benchmarks for other low resource languages. To this end, we propose a new benchmark named Korean balanced evaluation of significant tasks (KoBEST), which consists of five Korean-language downstream tasks. Professional Korean linguists designed the tasks that require advanced Korean linguistic knowledge. Moreover, our data is purely annotated by humans and thoroughly reviewed to guarantee high data quality. We also provide baseline models and human performance results. Our dataset is available on the Huggingface.

cs.CL↗

arXiv2021

The Untapped Potential of Off-the-Shelf Convolutional Neural Networks

Over recent years, a myriad of novel convolutional network architectures have been developed to advance state-of-the-art performance on challenging recognition tasks. As computational resources improve, a great deal of effort has been placed in efficiently scaling up existing designs and generating new architectures with Neural Architecture Search (NAS) algorithms. While network topology has proven to be a critical factor for model performance, we show that significant gains are being left on the table by keeping topology static at inference-time. Due to challenges such as scale variation, we should not expect static models configured to perform well across a training dataset to be optimally configured to handle all test data. In this work, we seek to expose the exciting potential of inference-time-dynamic models. By allowing just four layers to dynamically change configuration at inference-time, we show that existing off-the-shelf models like ResNet-50 are capable of over 95% accuracy on ImageNet. This level of performance currently exceeds that of models with over 20x more parameters and significantly more complex training procedures.

cs.CV↗

arXiv2020

Curvature Invariants for the Alcubierre and Nat\'ario Warp Drives

A process for using curvature invariants is applied to evaluate the metrics for the Alcubierre and the Natario warp drives at a constant velocity.Curvature invariants are independent of coordinate bases, so plotting these invariants will be free of coordinate mapping distortions. As a consequence, they provide a novel perspective into complex spacetimes such as warp drives. Warp drives are the theoretical solutions to Einstein's field equations that allow the possibility for faster-than-light (FTL) travel. While their mathematics is well established, the visualisation of such spacetimes is unexplored. This paper uses the methods of computing and plotting the warp drive curvature invariants to reveal these spacetimes. The warp drive parameters of velocity, skin depth and radius are varied individually and then plotted to see each parameter's unique effect on the surrounding curvature. For each warp drive, this research shows a safe harbor and how the shape function forms the warp bubble. The curvature plots for the constant velocity Natario warp drive do not contain a wake or a constant curvature indicating that these are unique features of the accelerating Natario warp drive.

gr-qc↗