arXiv subjects

Jiří Vyskočil

Publications and source records attributed to Jiří Vyskočil.

6 recordsLinked to original sources

arXiv2026

Concepts for Securing Agentic AI Coding and the Terok Environment

Agentic AI is a fascinating new tool for software development. It is a huge step forward compared to "conventional" AI assisted coding, which in turn was a considerable breakthrough earlier. AI support through LLMs is a young and very fast-moving field. The "conventional" (non-agentic) flavor became useful and productive in early 2025 (around 18 months ago) and the agentic flavor followed in fall 2025 (approximately 9 months ago). Besides all its benefits and potential, it also carries some fundamental risks for IT security. And the agentic approach added very severe risks while making others much more dangerous. With all the motivation to explore this fascinating new tool we should not ignore the risks but actively address them. We present (I) an assessment of the IT security risks, (II) a concept for mitigating them without breaking its benefits, and (III) an overview about an implementation of our concept. In this very dynamic field this is likely not the final and once-and-for-all answer to the identified issues but still a substantial step forward in responsible usage of Agentic AI for software development. It should also be a contribution to the community to allow early and eager evaluation of the potential of agentic AI for software development without actually suffering from its implied IT security risks.

cs.AI↗

arXiv2026

Nonlinear Matter Power Spectrum from relativistic $N$-body Simulations: $Λ_{\rm s}$CDM versus $Λ$CDM

We present relativistic $N$-body simulations of a $Λ_{\rm s}$CDM - sign-switching cosmological constant (CC) - scenario under general relativity and compare its nonlinear matter power spectrum to $Λ$CDM at ${z = 15,\,2,\,1,\,0}$, using best-fit parameters from Planck-only and a combined ''full'' dataset. During the AdS-like CC ($Λ_{\rm s}<0$) phase, prior to the transition redshift $z_\dagger$, reduced Hubble friction dynamically enhances the growth of perturbations; after the switch, with dS-like CC ($Λ_{\rm s}>0$), the larger late-time expansion rate partly suppresses, but does not erase, the earlier amplification. Consequently, the ratio $P_{Λ_{\rm s}\rm CDM}/P_{Λ\rm CDM}$ exhibits a pronounced, redshift-dependent shape feature: a crest peaking at ${\sim 20-25\%}$ around ${k \simeq 1-3\,h\,\mathrm{Mpc}^{-1}}$ near the transition, which then migrates to larger physical scales and persists to ${z = 0}$ as a robust ${\sim 15-20\%}$ uplift at ${k \simeq 0.6-1.0\,h\,\mathrm{Mpc}^{-1}}$. These wavenumbers correspond to group or poor-cluster environments and lie within the sensitivity range of weak lensing, galaxy-galaxy lensing, cluster counts, and tSZ power, providing a concrete, falsifiable target that cannot be mimicked by a scale-independent change in $σ_8$ or $S_8$. The timing (earlier for Planck-only, later for the full dataset) and the amplitude of the crest align with the ''cosmic noon'' epoch (${z \simeq 1-2}$), offering a gravitational prior for the observed peak in the cosmic star-formation rate.

astro-ph.CO↗

arXiv2024

Towards Zero-Shot Camera Trap Image Categorization

This paper describes the search for an alternative approach to the automatic categorization of camera trap images. First, we benchmark state-of-the-art classifiers using a single model for all images. Next, we evaluate methods combining MegaDetector with one or more classifiers and Segment Anything to assess their impact on reducing location-specific overfitting. Last, we propose and test two approaches using large language and foundational models, such as DINOv2, BioCLIP, BLIP, and ChatGPT, in a zero-shot scenario. Evaluation carried out on two publicly available datasets (WCT from New Zealand, CCT20 from the Southwestern US) and a private dataset (CEF from Central Europe) revealed that combining MegaDetector with two separate classifiers achieves the highest accuracy. This approach reduced the relative error of a single BEiTV2 classifier by approximately 42\% on CCT20, 48\% on CEF, and 75\% on WCT. Besides, as the background is removed, the error in terms of accuracy in new locations is reduced to half. The proposed zero-shot pipeline based on DINOv2 and FAISS achieved competitive results (1.0\% and 4.7\% smaller on CCT20, and CEF, respectively), which highlights the potential of zero-shot approaches for camera trap image categorization.

cs.CV↗

arXiv2019

Inverse Compton scattering from solid targets irradiated by ultra-short laser pulses in the $10^{22}-10^{23}\,\mathrm{W/cm^2}$ regime

Emission of high energy gamma rays via the non-linear inverse Compton scattering process (ICS) in interactions of ultra-intense laser pulses with thin solid foils is studied using particle-in-cell simulations. It is shown that the angular distribution of the ICS photons has a forward-oriented two-directional structure centred at an angle $\vartheta = \pm 30^\circ$, a value predicted by a theoretical model based on a standing wave approximation to the electromagnetic field in front of the target, which only increases at the highest intensities due to faster hole boring, which renders the approximation invalid. The conversion efficiency is shown to exhibit a super-linear increase with the driving pulse intensity. In comparison to emission via electron-nucleus bremsstrahlung, it is shown that the higher absorption, further enhanced by faster hole boring, in the targets with lower atomic number strongly favours the ICS process.

physics.plasm-ph↗

arXiv2016

Semantic Parsing of Mathematics by Context-based Learning from Aligned Corpora and Theorem Proving

We study methods for automated parsing of informal mathematical expressions into formal ones, a main prerequisite for deep computer understanding of informal mathematical texts. We propose a context-based parsing approach that combines efficient statistical learning of deep parse trees with their semantic pruning by type checking and large-theory automated theorem proving. We show that the methods very significantly improve on previous results in parsing theorems from the Flyspeck corpus.

cs.CL↗

arXiv2014

Machine Learner for Automated Reasoning 0.4 and 0.5

Machine Learner for Automated Reasoning (MaLARea) is a learning and reasoning system for proving in large formal libraries where thousands of theorems are available when attacking a new conjecture, and a large number of related problems and proofs can be used to learn specific theorem-proving knowledge. The last version of the system has by a large margin won the 2013 CASC LTB competition. This paper describes the motivation behind the methods used in MaLARea, discusses the general approach and the issues arising in evaluation of such system, and describes the Mizar@Turing100 and CASC'24 versions of MaLARea.

cs.LG↗