BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Models that prove their own correctness - Orr Paradise (UC Berkele
 y)
DTSTART:20250109T140000Z
DTEND:20250109T150000Z
UID:TALK224692@talks.cam.ac.uk
CONTACT:Tom Gur
DESCRIPTION:This talk introduces Self-Proving models\, a new class of mode
 ls that formally prove the correctness of their outputs via an Interactive
  Proof system. After reviewing some related literature\, I will formally d
 efine Self-Proving models and their per-input (worst-case) guarantees. I w
 ill then present algorithms for learning these models and explain how the 
 complexity of the proof system affects the complexity of the learning algo
 rithms. Finally\, I will show experiments where Self-Proving models are tr
 ained to compute the Greatest Common Divisor of two integers\, and to prov
 e the correctness of their results to a simple verifier.\n\nNo prior knowl
 edge of autoregressive models or Interactive Proofs will be assumed of the
  listener. This is a joint work with Noga Amit\, Shafi Goldwasser\, and Gu
 y Rothblum.
LOCATION:Computer Laboratory\, William Gates Building\, FW26
END:VEVENT
END:VCALENDAR