Rebecca Payne | Searcharxiv

arXiv2026

Veiled in Starlight: Impacts of Stellar Contamination on Retrievals of TRAPPIST-1f's Atmospheric Composition

The TRAPPIST-1 system offers seven terrestrial exoplanets with tight orbits and large radii ratios to the host star. If an atmosphere exists, transmission spectroscopy can be used to detect specific atmospheric features. Predictions of the atmospheric detectability of the TRAPPIST-1 planets prior to the launch of \textit{JWST} assumed pristine stellar surfaces. However, initial \textit{JWST} observations of the TRAPPIST-1 planets demonstrate that stellar contamination from unocculted active regions imparts significantly stronger spectral features than any planetary atmospheres. Here, we evaluate the atmospheric detectability of the habitable zone planet TRAPPIST-1f using atmospheric retrievals accounting for stellar contamination. We model a transmission spectrum given a CO$_2$-rich, habitable atmospheric model, and we include a "worst case" stellar contamination spectrum. We then perform atmospheric retrievals on simulated \textit{JWST} observations with MIRI LRS (5-15 \micron) and NIRSpec PRISM (0.6-5.3 \micron), assuming accurate starspot spectral models. We find that NIRSpec observations alone achieve similar results as MIRI and NIRSpec together. We find $\sim$10 transits obtains strong evidence ($B>150$) for CO$_2$, and $\sim$50 transits finds weak evidence ($B>3$) for CH$_4$. We could not retrieve evidence of H$_2$O with up to 100 simulated transits with both instruments. Many challenges remain to accurately account for stellar contamination for ultra-cool M-dwarfs in atmospheric retrievals, and our results show that, while evidence for a CO$_2$-rich atmosphere around TRAPPIST-1f can be found with a short \textit{JWST} program, other prominent atmospheric signatures can only be disentangled from strong stellar features with more observation time than previous studies have indicated.

astro-ph.EP↗

arXiv2025

Clinical knowledge in LLMs does not translate to human interactions

Global healthcare providers are exploring use of large language models (LLMs) to provide medical advice to the public. LLMs now achieve nearly perfect scores on medical licensing exams, but this does not necessarily translate to accurate performance in real-world settings. We tested if LLMs can assist members of the public in identifying underlying conditions and choosing a course of action (disposition) in ten medical scenarios in a controlled study with 1,298 participants. Participants were randomly assigned to receive assistance from an LLM (GPT-4o, Llama 3, Command R+) or a source of their choice (control). Tested alone, LLMs complete the scenarios accurately, correctly identifying conditions in 94.9% of cases and disposition in 56.3% on average. However, participants using the same LLMs identified relevant conditions in less than 34.5% of cases and disposition in less than 44.2%, both no better than the control group. We identify user interactions as a challenge to the deployment of LLMs for medical advice. Standard benchmarks for medical knowledge and simulated patient interactions do not predict the failures we find with human participants. Moving forward, we recommend systematic human user testing to evaluate interactive capabilities prior to public deployments in healthcare.

cs.HC↗