BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Yuan-Sen Ting (Ohio State University) DTSTART:20260312T183000Z DTEND:20260312T193000Z DTSTAMP:20260423T022733Z UID:nhetc/144 DESCRIPTION:Title: Expediting Astronomical Discovery with AI Agents\nby Yuan-Sen Ting (Oh io State University) as part of NHETC Seminar\n\n\nAbstract\nThe expansive \, interdisciplinary nature of astronomy\, combined with its open-access c ulture\, makes it an ideal testing ground for exploring how Large Language Models (LLMs) can accelerate scientific discovery. In this talk\, I will present our recent advances in applying LLMs as autonomous research agents . We demonstrate that AI agents now achieve gold-medal performance on Inte rnational Astronomy Olympiad problems\, and that multi-agent frameworks li ke Mephisto can conduct end-to-end galaxy spectral fitting—iteratively r efining physical models and accumulating knowledge through self-play\, app roaching human-like intuition and domain reasoning. However\, the Moravec paradox manifests clearly: tasks requiring abstract calculation may be eas ier for AI than seemingly simple perceptual tasks like chart reading and v isual reasoning\, which remain key bottlenecks. To address the cost barrie r at scale\, we developed open-source specialized models (AstroSage) that match frontier performance on astronomy Q&A at a fraction of the cost. I w ill conclude by reflecting on the epistemological implications—what coun ts as knowledge and understanding when AI agents can reason but not yet pe rceive the way scientists do\, and what this means for the future of astro nomical discovery.\n LOCATION:https://researchseminars.org/talk/nhetc/144/ END:VEVENT END:VCALENDAR