Hubble: a Model Suite to Advance the Study of LLM Memorization

ICLROral2026

Authors
Johnny Wei, Ameya Godbole, Mohammad Aflah Khan, Ryan Yixiang Wang, Xiaoyuan Zhu, James Flemings, Nitya Kashyap, Krishna P. Gummadi, Willie Neiswanger, Robin Jia
Affiliation
University of Southern California
Venue
ICLR 2026
Track
Oral

TL;DR

Hubble is a suite of paired LLMs (largest 8B), where the perturbed models are trained in the same way as standard models but with text (e.g. book passages, biographies, and test sets) inserted and designed to emulate key memorization risks.

Opening excerpt from the authors’ abstract. source

Read the paper

Topics

graph llm

← All ICLR 2026 Oral papers · Browse the whole archive