ContinuousBench : Can Differentially Private Synthetic Text Improve Capabilities?

Peihan Liu Columbia University, Lucas RosenblattNYUWeiwei KongGoogle ResearchNatalia PonomarevaGoogle ResearchGautam KamathUniversity of Waterloo& Vector InstituteRachel CummingsColumbia UniversityRoxana GeambasuColumbia UniversityYu GanGoogleLillian TsaiGoogle, Alex BieGoogle Research · 2026

Read ContinuousBench : Can Differentially Private Synthetic Text Improve Capabilities? by Peihan Liu Columbia University, Lucas RosenblattNYUWeiwei KongGoogle ResearchNatalia PonomarevaGoogle ResearchGautam KamathUniversity of Waterloo& Vector InstituteRachel CummingsColumbia UniversityRoxana GeambasuColumbia UniversityYu GanGoogleLillian TsaiGoogle, Alex BieGoogle Research (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-01849.txt.

Introduction

Liu and co-authors ask if private fake text can teach models new skills. Private text synthesis promises access to sealed corpora. But gains may just echo tasks models near-solve anyway. Past tests rely on such near-solved tasks. That makes real knowledge transfer hard to prove.

Their ContinuousBench test is rebuilt on a set cycle. Each release pairs a fresh corpus with new questions. The questions cannot be solved without that corpus. Yet hundreds of records support each tested fact. That design makes the task learnable under privacy rules. Two tracks cover fake creatures and fresh news. A shared harness trains models on fake text. Gains are then measured in a standard way.

Catalog record

CREATORPeihan Liu Columbia University, Lucas RosenblattNYUWeiwei KongGoogle ResearchNatalia PonomarevaGoogle ResearchGautam KamathUniversity of Waterloo& Vector InstituteRachel CummingsColumbia UniversityRoxana GeambasuColumbia UniversityYu GanGoogleLillian TsaiGoogle, Alex BieGoogle Research
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE140 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.01849

Browse the full library · About the Tower of Babel