SearcharxivSearch

arXiv subjects

Sebastian Kraus

Publications and source records attributed to Sebastian Kraus.

4 recordsLinked to original sources

Safeguards for Speech2Speech LLM-Assistants: A Case Study in Automotive Applications

Recent advances have introduced speech-to-speech (S2S) conversational assistants capable of producing natural-sounding interactions, including non-verbal cues like tonality and mood. In the automotive domain, this enables intuitive and humanlike in-car dialogue experiences. However, integrating these end-to-end assistants limits architectural options for programmable domain-specific safeguards. This paper discusses two implementation approaches for S2S guardrails: transcript-based and tool-based. Through an empirical evaluation, we demonstrate that both strategies are insufficient for industrial deployment in most cases due to prohibitive latency (delaying each answer by 0 to 1.4 seconds even for computationally cheap checks) and technical impediments (like potentially non-deterministic tool call behavior). Finally, we outline open challenges for S2S guardrails in the automotive context.

cs.AI

Effect of pop-up bike lanes on cycling in European cities

The bicycle is a low-cost means of transport linked to low risk of COVID-19 transmission. Governments have incentivized cycling by redistributing street space as part of their post-lockdown strategies. Here, we evaluate the impact of provisional bicycle infrastructure on cycling traffic in European cities. We scrape daily bicycle counts spanning over a decade from 736 bicycle counters in 106 European cities. We combine this with data on announced and completed pop-up bike lane road work projects. On average 11.5 kilometers of provisional pop-up bike lanes have been built per city. Each kilometer has increased cycling in a city by 0.6%. We calculate that the new infrastructure will generate $2.3 billion in health benefits per year, if cycling habits are sticky.

physics.soc-ph

Effects of thermal inversion induced air pollution on COVID-19

Air pollution is a threat to human health, in particular since it aggravates respiratory diseases. Early COVID-19 outbreaks in Wuhan, China and Lombardy, Italy coincided with high levels of air pollution drawing attention to a potential role of particulate matter and other pollutants in infections and more severe outcomes of the new lung disease. Both air pollution and COVID-19 outcomes are driven by human mobility and economic activity leading to spurious correlations in regression estimates. We use district-level panel data from Belgium, Brazil, Germany, Italy, the UK, and the US to estimate the impact of daily variation in air pollution levels on COVID-19 infections and deaths. Using random variation in air pollution generated by thermal inversions, we rule out that changes in mobility and economic activity are driving the results. We find that a 1%-increase in air pollution levels over the three preceding weeks leads to a 1.5% increase in weekly cases. A 1%-increase in air pollution over four weeks leads to 5.1% more COVID-19 deaths. These results indicate that short-term measures to reduce air pollution can help mitigate the health damages of the virus.

physics.soc-ph

Phase structure of the massive chiral Gross-Neveu model from Hartree-Fock

The phase diagram of the massive chiral Gross-Neveu model (the massive Nambu-Jona-Lasinio model in 1+1 dimensions) is constructed. In the large N limit, the Hartree-Fock approach can be used. We find numerically a chiral crystal phase separated from a massive Fermi gas phase by a 1st order transition. Using perturbation theory, we also construct the critical sheet where the homogeneous phase becomes unstable in a 2nd order transition. A tricritical curve is located. The phase diagram is mapped out as a function of fermion mass, chemical potential and temperature and compared with the one of the discrete chiral Gross-Neveu model. As a by-product, we illustrate the crystal structure of matter at zero temperature for various densities and fermion masses.

hep-th