AspenTech (United States)

Abstract Large language models hallucinate at rates that undermine reliability in high-stakes applications. Mitigation strategies based on repeated sampling or majority voting implicitly assume error independence across samples. We introduce Epistemic Field Theory (EFT), which predicts hallucination probability from multi-model consensus. EFT defines a consensus field σ ∈ [0, 1] over query space and derives the predictor P(H) = (1 − σ)·η, where η is a model-specific noise coefficient. Across 13,728 human-validated responses (Cohen’s κ = 0.87) from four frontier models in three professional domains, σ predicts hallucination with AUC = 0.787, outperforming majority voting (0.518), SelfCheck methods (0.358–0.377), and self-reported confidence (0.461). Hallucination counts exhibit systematic overdispersion (ρ = 1.50), with empirical majority-failure rates 2.96× higher than independence predicts. Epistemic grounding reduces hallucination rates but not error correlation, revealing frequency and structure as independent dimensions of the hallucination problem.

The field of metallic glasses has been an active area of research owing to the complex structure–property correlations and intricacies surrounding glass formation and relaxation. This review provides a thorough examination of significant works that elucidate the structure–property correlations of metallic glasses, derived from detailed atomistic simulations coupled with data-driven approaches. The review starts with the theoretical and fundamental framework for understanding important properties of metallic glasses such as transition temperatures, relaxation phenomena, the potential energy landscape, structural features such as soft spots and shear transformation zones, atomic stiffness and structural correlations. The need to understand these concepts for leveraging metallic glasses for a wide range of applications such as performance under tensile loading, viscoelastic properties, relaxation behavior and shock loading is also elucidated. Finally, the use of machine learning algorithms in predicting the properties of metallic glasses along with their applications, limitations and scope for future work is presented.Graphical abstract

A time-modulated LO (TM-LO) vector modulator (VM) architecture using a time domain approach for amplitude scaling and phase shifting received signals is presented. The TM-LO uses rail-to-rail LO waveforms generated from digitally synthesized blocks and pass-gate switches to perform the amplitude/phase control. A single element receiver achieves 0.2 dB RMS gain error and 1.4° RMS phase error with 5 bits of amplitude/phase resolution across a 360° range is implemented in a 65 nm CMOS process. Without time-modulation, the hardware is capable of 3-bits of resolution. The inherent digital nature of TM-LO architecture provides opportunity very compact front-ends suitable for large arrays and lower voltage technologies. Four TM-LO chips were used to create a beamforming receiver

To evaluate the safety and efficacy of Fortelyzin in the performance of staged reperfusion therapy for acute ischemic stroke (intravenous thrombolytic therapy plus mechanical thrombectomy) in anterior circulation within the FORTA RF multicenter pilot study. The study included 72 patients with acute ischemic stroke in anterior circulation, who underwent staged reperfusion therapy in four vascular centers of the Russian Federation from December 2019 to January 2023. The mean time from illness onset to hospitalization was 94.5 minutes in the Fortelyzin group and 97.2 min in the Actilyse group (p=0.78). The time from the moment of hospitalization to the admission of the patient to the X-ray operating room was significantly lower in the Fortelyzin group (p=0.002). The incidence of symptomatic hemorrhagic transformations in the Fortelyzin group was 6%, in the Actilyse group - 8% (p=0.75). A favorable functional outcome in the first group was observed in 47% of patients, in the control group in 42% (p=0.66). Mortality in both groups did not differ significantly and amounted to 22% and 25%, respectively. The first results of the FORTA RF multicenter study demonstrate the safety and efficacy of Fortelyzin in staged reperfusion therapy compared to Actilyse.

BackgroundThe Symptoms of Infection with Coronavirus-19 (SIC) is a 30-item patient-reported outcome (PRO) measure scored by body system composites to assess signs/symptoms of coronavirus disease 2019 (COVID-19). In addition to cross-sectional and longitudinal psychometric evaluations, qualitative exit interviews were conducted to support the content validity of the SIC.MethodsIn a cross-sectional study, adults diagnosed with COVID-19 in the United States completed the web-based SIC and additional PRO measures. A subset was invited to participate in phone-based exit interviews. Longitudinal psychometric properties were assessed in ENSEMBLE2, a multinational, randomized, double-blind, placebo-controlled, phase 3 trial of the Ad26.COV2.S COVID-19 vaccine. Psychometric properties evaluated included structure, scoring, reliability, construct validity, discriminating ability, responsiveness, and meaningful change thresholds of SIC items and composite scores.ResultsIn the cross-sectional study, 152 participants completed the SIC (mean age, 51.0 ± 18.6 years) and 20 completed follow-up interviews. Fatigue (77.6%), feeling unwell (65.8%), and cough (60.5%) were symptoms most frequently reported. SIC inter-item correlations were all positive and mostly moderate (r ≥ 0.3) and statistically significant. SIC items and Patient-Reported Outcomes Measurement Information System-29 (PROMIS-29) scores correlated as hypothesized (all r ≥ 0.32). Internal consistency reliabilities of all SIC composite scores were satisfactory (Cronbach’s alpha, 0.69–0.91). SIC composite scores correlated moderately (r = 0.30–0.49) to strongly (r ≥ 0.50) with PROMIS-29 scores and Patient Global Impression of Severity (PGIS) ratings (all P < 0.01). A variety of signs/symptoms were cited in exit interviews, and participants considered the SIC straightforward, comprehensive, and easy to use. From ENSEMBLE2, 183 participants with laboratory-confirmed moderate to severe/critical COVID-19 were included (51.5 ± 14.8 years). Strong test-retest reliabilities were observed for most SIC composite scores (intraclass correlations ≥ 0.60). Statistically significant differences across PGIS severity levels were found for all but 1 composite score, supporting known-groups validity. All SIC composite scores demonstrated responsiveness based on changes in PGIS.ConclusionsThe psychometric evaluations provided strong evidence for the reliability and validity of the SIC for measuring COVID-19 symptoms, supporting its use in vaccine and treatment trials. In exit interviews, participants described a broad range of signs/symptoms consistent with previous research, further supporting the content validity and format of the SIC.

The Ricker and Gabor wavelets and their Hilbert transforms are widely used in many physical/engineering fields. In seismology, they are commonly utilized to describe source functions involved in the computation of synthetic seismograms in acoustic or elastic media, or inversely, the extraction of source wavelets from real seismic traces. We provide analytical formulae of the Ricker and Gabor wavelets’ Hilbert transforms, yielding to analytic expressions of the envelope, instantaneous frequency, and phase attributes of these wavelets. We analyze the plausibility of these wavelets by revealing the regions of positive instantaneous frequency and nonzero envelope.

Summary Well inflow in reservoirs with a random permeability is studied within stochastic moment equations (SME) approach. This method allows uncertainty quantification in well production without the need for multivariate calculations and provides acceptable accuracy. In formations with random permeability well productivity index can be represented as a random variable and the aim of this study to calculate its statistical moments. The knowledge of productivity index variance can be useful in history matching process, since it determines the confidence interval for a reasonable change in well productivity. It is also demonstrated that the common expression for well productivity index can lead to errors in well inflow numerical modeling in reservoirs with significant permeability fluctuations. The solution of SME allows to modify this common expression and to improve the accuracy of well modeling.