arXiv subjects

Jonathan Ng

Publications and source records attributed to Jonathan Ng.

14 recordsLinked to original sources

arXiv2026

Fingerprinting All AI Cluster I/O Without Mutually Trusted Processors

In preparation for potential international agreements on artificial intelligence, the development of verification infrastructure for AI data centres is vital. We propose a method for cryptographically committing all information entering and leaving a data centre: Hashes are computed by network taps placed on all the information-carrying wires between the cluster and the outside world, enabling an auditor to retroactively challenge the preimage data to be sent to a privacy-preserving verification facility performing compliance checks. Our goal is to make it infeasible to covertly exfiltrate the results of undisclosed workloads in the cluster through the tapped wires. To this end, we specify the architecture of a ``Secure Gateway Device'', which handles the erasure of covert channels that post-hoc verification on hashed data cannot address: analogue and timing side-channels, as well as steganography in network protocol headers. The architecture eliminates the need for any processors trusted by both the Prover and the Verifier, leveraging passive optical fibre splitters and coin-flip protocols for random number generation where needed. We expect development costs of a demonstration device to be roughly equivalent to the cost of a small team of engineers for a few months, with a comparatively small bill of materials.

cs.CR↗

arXiv2026

Phoenix-VL 1.5 Medium Technical Report

We introduce Phoenix-VL 1.5 Medium, a 123B-parameter natively multimodal and multilingual foundation model, adapted to regional languages and the Singapore context. Developed as a sovereign AI asset, it demonstrates that deep domain adaptation can be achieved with minimal degradation to broad-spectrum intelligence and alignment. Continued pretraining was performed on Mistral Medium 3.1 using a localized 1-trillion tokens multimodal corpus, followed by a 250-billion tokens long-context extension phase. Subsequent post-training incorporated a novel human-annotated Singapore multimodal dataset and curated textual corpus on Singapore culture, knowledge, and legislation, totaling 22-billion tokens. An additional 5 billion tokens of model alignment was performed through Online Direct Preference Optimization. Phoenix-VL 1.5 Medium achieves state-of-the-art performance for its size on Singapore multimodal, legal, and government policy benchmarks while remaining globally competitive on general multimodal intelligence, multilingual, and STEM benchmarks. We also introduce a novel evaluation suite encompassing localized knowledge benchmarks and an institutionally aligned model behavior and safety framework. We report the data curation principles, training methodology, and highlight benchmark and inference performance.

cs.CL↗

arXiv2024

Interaction of the Prominence Plasma within the Magnetic Cloud of an ICME with the Earth's Bow Shock

The magnetic cloud within an interplanetary coronal mass ejection (ICME) is characterized by high magnetic field intensities. In this study, we investigate the interaction of a magnetic cloud carrying a density structure with the Earth's bow shock during the ICME event on 24 April 2023. Elevated abundances of cold protons and heavier ions, namely alpha particles and singly charged helium ions, associated with the prominence plasma are observed within this structure. The plasma downstream of the bow shock exhibits an irregular compression pattern which could be due to the presence of heavy ions. Heavy ions carry a significant fraction of the upstream flow energy; however, due to their different charge per mass ratio and rigidity, they are less scattered by the electromagnetic and electrostatic waves at the shock. We find that downstream of the shock, while the thermal ion energy is only a small fraction of the background magnetic energy density, nevertheless increased ion fluxes reduce the characteristic wave speeds in the that region. As such, we observe a transition state of an unstable bow shock layer across which the plasma flow is super Alfv\'enic in both upstream and downstream regions. Our findings help with understanding the intense space weather impacts of such events.

physics.space-ph↗

arXiv2024

Catastrophic Cyber Capabilities Benchmark (3CB): Robustly Evaluating LLM Agent Cyber Offense Capabilities

LLM agents have the potential to revolutionize defensive cyber operations, but their offensive capabilities are not yet fully understood. To prepare for emerging threats, model developers and governments are evaluating the cyber capabilities of foundation models. However, these assessments often lack transparency and a comprehensive focus on offensive capabilities. In response, we introduce the Catastrophic Cyber Capabilities Benchmark (3CB), a novel framework designed to rigorously assess the real-world offensive capabilities of LLM agents. Our evaluation of modern LLMs on 3CB reveals that frontier models, such as GPT-4o and Claude 3.5 Sonnet, can perform offensive tasks such as reconnaissance and exploitation across domains ranging from binary analysis to web technologies. Conversely, smaller open-source models exhibit limited offensive capabilities. Our software solution and the corresponding benchmark provides a critical tool to reduce the gap between rapidly improving capabilities and robustness of cyber offense evaluations, aiding in the safer deployment and regulation of these powerful technologies.

cs.CR↗

arXiv2024

Impact of the Out-of-Plane Flow Shear on Magnetic Reconnection at the Flanks of Earth's Magnetopause

Magnetic reconnection changes the magnetic field topology and facilitates the energy and particle exchange at magnetospheric boundaries such as the Earth's magnetopause. The flow shear perpendicular to the reconnecting plane prevails at the flank magnetopause under southward interplanetary magnetic field (IMF) conditions. However, the effect of the out-of-plane flow shear on asymmetric reconnection is an open question. In this study, we utilize kinetic simulations to investigate the impact of the out-of-plane flow shear on asymmetric reconnection. By systematically varying the flow shear strength, we analyze the flow shear effects on the reconnection rate, the diffusion region structure, and the energy conversion rate. We find that the reconnection rate increases with the upstream out-of-plane flow shear, and for the same upstream conditions, it is higher at the dusk side than at the dawn side. The diffusion region is squeezed in the outflow direction due to magnetic pressure which is proportional to the square of the Alfv\'en Mach number of the shear flow. The out-of-plane flow shear increases the energy conversion rate J \cdot E', and for the same upstream conditions, the magnitude of J \cdot E' is larger at the dusk side than at the dawn side. This study reveals that out-of-plane flow shear not only enhances the reconnection rate but also significantly boosts energy conversion, with more pronounced effects on the dusk-side flank than on the dawn-side flank. These insights pave the way for better understanding the solar wind-magnetosphere interactions.

physics.space-ph↗

arXiv2024

Earth's Alfv\'en wings driven by the April 2023 Coronal Mass Ejection

We report a rare regime of Earth's magnetosphere interaction with sub-Alfv\'enic solar wind in which the windsock-like magnetosphere transforms into one with Alfv\'en wings. In the magnetic cloud of a Coronal Mass Ejection (CME) on April 24, 2023, NASA's Magnetospheric Multiscale mission distinguishes the following features: (1) unshocked and accelerated cold CME plasma coming directly against Earth's dayside magnetosphere; (2) dynamical wing filaments representing new channels of magnetic connection between the magnetosphere and foot points of the Sun's erupted flux rope; (3) cold CME ions observed with energized counter-streaming electrons, evidence of CME plasma captured due to reconnection between magnetic-cloud and Alfv\'en-wing field lines. The reported measurements advance our knowledge of CME interaction with planetary magnetospheres, and open new opportunities to understand how sub-Alfv\'enic plasma flows impact astrophysical bodies such as Mercury, moons of Jupiter, and exoplanets close to their host stars.

physics.space-ph↗

arXiv2023

Do the Rewards Justify the Means? Measuring Trade-Offs Between Rewards and Ethical Behavior in the MACHIAVELLI Benchmark

Artificial agents have traditionally been trained to maximize reward, which may incentivize power-seeking and deception, analogous to how next-token prediction in language models (LMs) may incentivize toxicity. So do agents naturally learn to be Machiavellian? And how do we measure these behaviors in general-purpose models such as GPT-4? Towards answering these questions, we introduce MACHIAVELLI, a benchmark of 134 Choose-Your-Own-Adventure games containing over half a million rich, diverse scenarios that center on social decision-making. Scenario labeling is automated with LMs, which are more performant than human annotators. We mathematize dozens of harmful behaviors and use our annotations to evaluate agents' tendencies to be power-seeking, cause disutility, and commit ethical violations. We observe some tension between maximizing reward and behaving ethically. To improve this trade-off, we investigate LM-based methods to steer agents' towards less harmful behaviors. Our results show that agents can both act competently and morally, so concrete progress can currently be made in machine ethics--designing agents that are Pareto improvements in both safety and capabilities.

cs.LG↗

arXiv2021

Magnetopause reconnection and indents induced by foreshock turbulence

Based on global hybrid simulation results, we predict that foreshock turbulence can reach the magnetopause and lead to reconnection as well as Earth-sized indents. Both the interplanetary magnetic field (IMF) and solar wind are constant in our simulation, and hence all dynamics are generated by foreshock instabilities. The IMF in the simulation is mostly Sun-Earth aligned with a weak northward and zero dawn-dusk component, such that subsolar magnetopause reconnection is not expected without foreshock turbulence modifying the magnetosheath fields. We show a reconnection example to illustrate that the turbulence can create large magnetic shear angles across the magnetopause to induce local bursty reconnection. Magnetopause reconnection and indents developed from the impact of foreshock turbulence can potentially contribute to dayside loss of planetary plasmas.

physics.space-ph↗

arXiv2020

Ion-scale current structures in Short Large-Amplitude Magnetic Structures

We investigate electric current structures in Short Large-Amplitude Magnetic Structures (SLAMS) in the terrestrial ion foreshock region observed by the Magnetospheric Multiscale mission. The structures with intense currents (|J|~1 {\mu}A/m^2) have scale lengths comparable to the local ion inertial length (di). One current structure type is a current sheet due to the magnetic field rotation of the SLAMS, and a subset of these current sheets can exhibit reconnection features including the electron outflow jet and X-line-type magnetic topology. The di-scale current sheet near the edge of a SLAMS propagates much more slowly than the overall SLAMS, suggesting that it may result from compression. The current structures also exist as magnetosonic whistler waves with fci < f < flh, where fci and flh are the ion cyclotron frequency and the lower-hybrid frequency, respectively. The field rotations in the current sheets and whistler waves generate comparable |J| and energy conversion rates. Electron heating is clearly observed in one whistler packet embedded in a larger-scale current sheet of the SLAMS, where the parallel electric field and the curvature drift opposite to the electric field energize electrons. The results give insight about the thin current structure generation and energy conversion at thin current structures in the shock transition region.

physics.space-ph↗

arXiv2019

Three-Dimensional Stability of Current Sheets Supported by Electron Pressure Anisotropy

The stability of electron current sheets embedded within the reconnection exhaust is studied with a 3D fully kinetic particle-in-cell simulation. The electron current layers studied here form self-consistently in a reconnection regime with a moderate guide field, are supported by electron pressure anisotropy with the pressure component parallel to the magnetic field direction larger than the perpendicular components, and extend well beyond electron kinetic scales. In 3D, in addition to drift instabilities common to nearly all reconnection exhausts, the regime considered also exhibits an electromagnetic instability driven by the electron pressure anisotropy. While the fluctuations modulate the current density on small scales, they do not break apart the general structure of the extended electron current layers. The elongated current sheets should therefore persist long enough to be observed both in space observations and in laboratory experiments.

physics.plasm-ph↗

arXiv2019

Exact and Locally Implicit Source Term Solvers for Multifluid-Maxwell Systems

Recently, a family of models that couple multifluid systems to the full Maxwell equations draw a lot of attention in laboratory, space, and astrophysical plasma modeling. These models are more complete descriptions of the plasma than reduced models like magnetohydrodynamic (MHD) since they naturally retain non-ideal effects like electron inertia, Hall term, pressure anisotropy/nongyrotropy, etc. One obstacle to broader application of these model is that an explicit treatment of their source terms leads to the need to resolve rapid kinetic processes like plasma oscillation and electron cyclotron motion, even when they are not important. In this paper, we suggest two ways to address this issue. First, we derive the analytic forms solutions to the source update equations, which can be implemented as a practical, but less generic solver. We then develop a time-centered, locally implicit algorithm to update the source terms, allowing stepping over the fast kinetic time-scales. For a plasma with $S$ species, the locally implict algorithm involves inverting a local $3S+3$ matrix only, thus is very efficient. The performance can be further elevated by using the direct update formulas to skip null calculations. Benchmarks illustrated the exact energy-conservation of the locally implicit solver, as well as its efficiency and robustness for both small-scale, idealized problems and large-scale, complex systems. The locally implicit algorithm can be also easily extended to include other local sources, like collisions and ionization, which are difficult to solve analytically.

physics.comp-ph↗

arXiv2018

Using the maximum entropy distribution to describe electrons in reconnecting current sheets

Particle distributions in weakly collisional environments such as the magnetosphere have been observed to show deviations from the Maxwellian distribution. These can often be reproduced in kinetic simulations, but fluid models, which are used in global simulations of the magnetosphere, do not necessarily capture any of this. We apply the maximum entropy fluid closure of Levermore, which leads to well posed moment equations, to reconstruct particle distributions from a kinetic simulation in a reconnection region. Our results show that without information other than the moments, the model can reproduce the general structure of the distributions but not all of the finer details. The advantages of the closure over the traditional Grad closure are also discussed.

physics.space-ph↗

arXiv2016

The role of guide field in magnetic reconnection driven by island coalescence

A number of studies have considered how the rate of magnetic reconnection scales in large and weakly collisional systems by the modelling of long reconnecting current sheets. However, this set-up neglects both the formation of the current sheet and the coupling between the diffusion region and a larger system that supplies the magnetic flux. Recent studies of magnetic island merging, which naturally include these features, have found that ion kinetic physics is crucial to describe the reconnection rate and global evolution of such systems. In this paper, the effect of a guide field on reconnection during island merging is considered. In contrast to the earlier current sheet studies, we identify a limited range of guide fields for which the reconnection rate, outflow velocity, and pile-up magnetic field increase in magnitude as the guide field increases. The Hall-MHD fluid model is found to reproduce kinetic reconnection rates only for a sufficiently strong guide field, for which ion inertia breaks the frozen-in condition and the outflow becomes Alfvenic in the kinetic system. The merging of large islands occurs on a longer timescale in the zero guide field limit, which may in part be due to a mirror-like instability that occurs upstream of the reconnection region.

physics.plasm-ph↗

arXiv2015

The island coalescence problem: scaling of reconnection in extended fluid models including higher-order moments

As modeling of collisionless magnetic reconnection in most space plasmas with realistic parameters is beyond the capability of today's simulations, due to the separation between global and kinetic length scales, it is important to establish scaling relations in model problems so as to extrapolate to realistic scales. Recently, large scale particle-in-cell (PIC) simulations of island coalescence have shown that the time averaged reconnection rate decreases with system size, while fluid systems at such large scales in the Hall regime have not been studied. Here we perform the complementary resistive MHD, Hall MHD and two fluid simulations using a ten-moment model with the same geometry. In contrast to the standard Harris sheet reconnection problem, Hall MHD is insufficient to capture the physics of the reconnection region. Additionally, motivated by the results of a recent set of hybrid simulations which show the importance of ion kinetics in this geometry, we evaluate the efficacy of the ten-moment model in reproducing such results.

physics.plasm-ph↗