BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:www.tcs.tifr.res.in/event/1740
DTSTAMP:20260702T052133Z
SUMMARY:Just Autocomplete??
DESCRIPTION:Speaker: Nishant P. Das\n\nAbstract: \nFor a time\, it was comm
 on to dismiss large language models (LLMs) as "just autocomplete" and ther
 efore fundamentally incapable of sophisticated reasoning. Yet modern LLMs 
 exhibit capabilities that would have seemed impossible only a few years ag
 o\, the latest of which is disproving a longstanding conjecture of Erdős 
 on the unit distance problem. One might attribute this progress to larger 
 models and larger training corpora\, but this is far from the whole story:
  after one has already scraped most of the available high-quality text on 
 the internet\, obtaining substantially more training data becomes challeng
 ing. Instead\, much of the recent progress has come from innovations in ho
 w LLMs are trained. In this talk\, we will trace the evolution of post-tra
 ining and inference procedures (RLHF\, DPO\, RLVR\, test-time compute\, et
 c.) that have transformed LLMs into much more than mere autocomplete.\n \
 n
URL:https://www.tcs.tifr.res.in/web/events/1740
DTSTART;TZID=Asia/Kolkata:20260703T160000
DTEND;TZID=Asia/Kolkata:20260703T170000
LOCATION:A-201 (STCS Seminar Room)
END:VEVENT
END:VCALENDAR