BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260421T090513Z
LOCATION:Bldg. 8 - Room B 102
DTSTART;TZID=Europe/Stockholm:20260701T093000
DTEND;TZID=Europe/Stockholm:20260701T100000
UID:submissions.pasc-conference.org_PASC26_sess157_msa207@linklings.com
SUMMARY:Semantic Provenance for Trustworthy Agentic Workflows in Materials
  Science
DESCRIPTION:Edan Bainglass, Xing Wang, Alexander Goscinski, Julian Geiger,
  and Giovanni Pizzi (Paul Scherrer Institute)\n\nAgentic AI systems, auton
 omous agents capable of planning simulations, invoking computational tools
 , and reasoning over results, offer new opportunities for accelerating dis
 covery in materials science. However, ensuring reproducibility of research
  endeavors remains a key challenge when autonomous systems orchestrate lar
 ge computational campaigns.\n\nWe present a graph-native protocol designed
  to support trustworthy agent-driven scientific workflows by separating sc
 ientific intent from execution infrastructure. The protocol organizes work
 flow execution into three complementary representations: a workflow graph 
 describing task structure and data dependencies, a provenance graph record
 ing the complete execution history, and a knowledge graph capturing semant
 ic relationships between scientific quantities and results.\n\nTogether, t
 hese representations provide machine-actionable context that enables auton
 omous agents to reason about physical quantities and retrieve the computat
 ional evidence underlying scientific conclusions. To demonstrate portabili
 ty and reproducibility, we execute a single workflow specification across 
 eleven heterogeneous workflow engines, including Airflow, Dask, Parsl, and
  Jobflow and spanning diverse HPC environments. The provenance graph is st
 ored using AiiDA’s provenance implementation. Despite differences in runti
 me infrastructure, all systems produce identical provenance graphs from th
 e same workflow definition.\n\nThis approach enables AI-driven computation
 al campaigns that remain transparent and reproducible, providing a foundat
 ion for generating FAIR, machine-actionable datasets and supporting trustw
 orthy AI-assisted scientific discovery.\n\nDomain: Chemistry and Materials
 , Climate, Weather, and Earth Sciences, Engineering, Physics\n\nSession Ch
 air: Jan Janssen (Max Planck Institute for Sustainable Materials)\n\n
END:VEVENT
END:VCALENDAR