CRED: Counterfactual Reasoning and Environment Design for Active Preference Learning

Yi-Shiuan Tung, Gyanig Kumar, Wei Jiang, Bradley Hayes, Alessandro Roncone · 2026

Read CRED: Counterfactual Reasoning and Environment Design for Active Preference Learning by Yi-Shiuan Tung, Gyanig Kumar, Wei Jiang, Bradley Hayes, Alessandro Roncone (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-2603-08531.txt.

Description

Yi-Shiuan Tung and colleagues teach robots to learn human tastes by inventing fresh test scenes. Old preference learning reuses fixed clips or past runs, which starves it of varied, telling comparisons. Their CRED method designs new settings and poses what-if reward pairs, beating rival methods on reward accuracy and sample thrift in tests and a user study.

Catalog record

CREATORYi-Shiuan Tung, Gyanig Kumar, Wei Jiang, Bradley Hayes, Alessandro Roncone
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE39034
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.08531

Browse the full library · About the Tower of Babel