arXiv subjects

Benjamin Berczi

Publications and source records attributed to Benjamin Berczi.

4 recordsLinked to original sources

arXiv2026

You Are What You Read: Misalignment via In-Context Persona Induction

Broad misalignment has been produced by finetuning on narrow data, harmful or benign, and in context only by demonstrations of the undesirable behaviour itself. We show that benign data suffices in context, with no finetuning and no demonstration of harmful behaviour in the prompt. Biographical facts that converge on a single figure, placed in a model's context as ordinary conversational turns, lead it to answer as that figure on questions the facts never touch. We call this persona induction. Across nine personas and thirteen models, identity adoption rises sigmoidally with the number of facts and crosses 50% within 3 to 10 of them. Misalignment then tracks which figure is described. Harmless personas reach full adoption with near-zero misalignment, while harmful ones voice their characteristic views on unrelated questions, at rates up to 80%. A formatting instruction can gate when the persona activates. Because each fact is individually benign, accumulated biographical context is flagged by content filters on 3% of inputs against 24-33% for an equivalent direct instruction.

cs.CL↗

arXiv2024

Quantum correlations in a gravitational collapse simulation with SpheriCo.jl

We report on work using a newly developed code, SpheriCo.jl, that computes the gravitational collapse of a spherical scalar field, where the scalar can be either a classical field, or a quantum field operator. By utilising summation-by-parts methods for the numerical derivatives we are able to simulate the collapse longer than was possible previously due to enhanced numerical stability. We present a suite of tests for the code that tests its accuracy and stability, both for the classical and quantum fields. We are able to observe critical behavior of gravitational collapse for the classical setup, in agreement with expected results. The code is also used to compute two-point correlation functions, with results that hint at a non-trivial correlation across the horizon of Hawking quanta.

gr-qc↗

arXiv2021

Gravitational collapse of quantum fields and Choptuik scaling

Gravitational collapse into a black hole has been extensively studied with classical sources. We develop a new formalism to simulate quantum fields forming a black hole. By choosing a convenient coherent state, this formalism taps into well-established techniques used for classical collapse and adds on the evolution of the mode functions of the quantum field operator. Divergences are regularized with the cosmological constant and Pauli-Villars fields. Using a massless spherically symmetric scalar field as an example, we demonstrate the effectiveness of the formalism by reproducing some classical results in gravitational collapse, and identifying the difference due to the quantum effects. We also find that Choptuik scaling in critical collapse survives in the semiclassical simulation, and furthermore the quantum deviation from the classical Choptuik scaling decreases when the system approaches the critical point.

hep-th↗

arXiv2020

Gravitational collapse with quantum fields

Gravitational collapse into a black hole has been extensively studied with classical sources. We develop a new formalism to simulate quantum fields forming a black hole. This formalism utilizes well-established techniques used for classical collapse by choosing a convenient coherent state, and simulates the matter fields quantum mechanically. Divergences are regularized with the cosmological constant and Pauli-Villars fields. Using a massless spherically symmetric scalar field as an example, we demonstrate the effectiveness of the formalism by reproducing some classical results in gravitational collapse, and identifying the difference due to quantum effects.

gr-qc↗