BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260202T201248Z
LOCATION:Second Floor Atrium
DTSTART;TZID=America/Chicago:20251120T080000
DTEND;TZID=America/Chicago:20251120T170000
UID:submissions.supercomputing.org_SC25_sess533_post299@linklings.com
SUMMARY:Understanding LLM Behavior on HPC Data via Mechanistic Interpretab
 ility
DESCRIPTION:Md Mahbubur Rahman (Iowa State University), Arjun Guha (Northe
 astern University), and Harshitha Menon (Lawrence Livermore National Labor
 atory (LLNL))\n\nLarge language models (LLMs) are increasingly used in HPC
  for tasks like code generation and analysis, but their internal reasoning
  remains opaque. To address this, we study three tasks—OpenMP code complet
 ion, data race detection, and OMP code generation—using mechanistic interp
 retability. Sparse autoencoder ablations reveal causal features, function 
 vector injection improves zero-shot predictions and direction vector shift
 s the model's output toward a desired behavior or style, even without expl
 icitly stating it in the prompt. These methods expose and influence LLM be
 havior in HPC contexts.\n\nTag: Research & ACM SRC Posters\n\nRegistration
  Category: Technical Program Reg Pass\n\nSession Chairs: Kento Sato (RIKEN
  Center for Computational Science (R-CCS)); Chris Schlipalius (Pawsey Supe
 rcomputing Research Centre; Commonwealth Scientific and Industrial Researc
 h Organisation (CSIRO), Australia); and Anja Gerbes (Georg-August-Universi
 tät Göttingen)\n\n
END:VEVENT
END:VCALENDAR