We present spectro-photometric griz colors for 189 near-Earth objects (NEOs) collected by the Mission Accessible Near-Earth Object Survey (MANOS). Data acquisition involved nonsimultaneous multiband exposures; thus, particular attention was given to the influence of rotational lightcurves on the derived colors. We show that colors measured without accounting for lightcurve variations can significantly influence results for individual objects and potentially introduce systematic offsets for ensemble studies. Color-based taxonomic classifications were used to investigate the distribution of spectral types. Our results were combined with other visible wavelength surveys to highlight a previously reported change in the observed taxonomic distribution of NEOs as a function of size, namely a decrease in S complex and an increase in X complex objects with increasing absolute magnitude. Plausibility arguments are given to suggest that Main Belt source region, thermal modification, discovery bias, tidal resurfacing, regolith grain size, and impact shock darkening are unlikely explanations for this size-dependent trend. Consistent with recent NEO population models and work on the connection between meteorites and young asteroid families in the Main Belt, this trend is best explained by a compositional gradient in the NEO population. In particular, the observed abundance of S complex or ordinary chondrite-like NEOs decreases by a factor of 2, from similar to 65% of the population at kilometer scales down to a third at sizes less than or similar to 50 m. This result has implications for understanding the initial pre-impact population of meteorite parent bodies prior to atmospheric filtering. Furthermore, this will have implications for probabilistic impact risk assessment models.
Virtual Reality (VR) is transforming education by providing immersive and interactive learning experiences. The integration of Artificial Intelligence (AI) augments these capabilities by providing adaptive, personalized support tailored to learners’ needs. This paper presents a conceptual framework for integrating advanced technologies into VR-based technical education, using Omlox, an industrial localization standard, as an illustrative example. Furthermore, we demonstrate how incorporating a Pedagogical Agent powered by Conversational AI enables real-time guidance, context-aware feedback, and interactive learning, fostering engagement, conceptual understanding, and skill development.
Estimates of how long the supply of human-written text will last multiply a raw token count by two factors: the fraction that survives quality filtering, and the number of times a surviving token can usefully be repeated. These two factors are not independent. On a fixed pool, filtering to the best fraction f does not save tokens; it multiplies the number of passes over the data by 1/f, and repeated tokens are worth less than fresh ones. We measure the resulting trade-off directly. Starting from a fixed 24-million-token pool of FineWeb and the released FineWeb-Edu quality classifier, we build eleven nested corpora—five quality-filtered at retention rates from 1 to 1 16 , four random subsets matched to those sizes, and two that keep everything but oversample the high-scoring slice—and train small transformers on all of them at matched compute, evaluating on raw web text, on high-scoring web text, and on WikiText-103. Three findings clear a measured run-to-run noise band of 0.055 nats per token. First, the sign of the filtering effect depends on the target: filtering never helps on the distribution the pool was drawn from, costing up to 0.44 nats, while on the two high-quality targets a retention rate chosen for the budget is at worst harmless and sometimes helps. Second, that rate loosens as the budget grows. Over a single doubling of the budget, the top-sixteenth corpus goes from marginally better than the unfiltered pool on WikiText-103 (0.061 nats, at the edge of the noise band) to clearly worse (0.298 nats), purely because it is traversed 8.2 times instead of 4.1. Third, at aggressive retention rates, deleting the low-scoring documents is worse than keeping them and oversampling the high-scoring ones—by up to 0.39 nats, on every target and at every budget—while at mild retention rates the two instruments are indistinguishable. We also report a negative result: two budget points per corpus are not enough to identify a repetition-decay constant, and we give the interval to show how badly. The experiments are four orders of magnitude below frontier scale, and we are explicit about what that does and does not license.
Der europaweite Ausbau von Rechenzentren eröffnet ein erhebliches Potenzial für die Systemstabilisierung im Energiesektor. Da Rechenprozesse ortsunabhängig durchgeführt werden können, besitzen Rechenzentren eine besondere Fähigkeit zur räumlichen Lastverschiebung zwischen Synchrongebieten. Diese Form der Flexibilisierung ermöglicht es, den Energieverbrauch dort zu reduzieren, wo Unterdeckung herrscht, und gleichzeitig in Regionen mit Energieüberdeckung zu erhöhen. Damit können Rechenzentren simultan zur Entlastung zweier Stromsysteme beitragen. Analysen zu Deutschland, Großbritannien und Finnland zeigen, dass eine bewusste Verschiebung von Rechenleistung in 94
Prolonged turbulence can result in the experience of kinetosis (motion sickness) among helicopter passengers, thereby negatively impacting the ride comfort of the flight. To address this issue, a comfort mode, analogous to that employed in the automotive industry, can be implemented for the helicopter flight control system. In order to design the comfort mode, a closed-loop reference model of the helicopter is derived that leads to minimal kinetosis at given turbulence intensities. Adaptive control laws from the model reference adaptive control framework are employed and evaluated to modify nominal closed-loop behavior in accordance with the derived reference model when the comfort mode is engaged. The efficacy of adaptive control laws in reducing the severity of kinetosis is demonstrated through simulations. These control laws compensate for the negative effects of turbulence, thereby contributing to a more comfortable flight experience.
