Reasoning Up the Instruction Ladder for Controllable Language Models
Published in Findings of the Association for Computational Linguistics: ACL 2026, 2026
We formulate instruction hierarchy resolution as a reasoning task and introduce VerIH, a verifiable training dataset for improving instruction prioritization and robustness.
Recommended citation: Zishuo Zheng, Vidhisha Balachandran, Chan Young Park, Faeze Brahman, and Sachin Kumar. 2026. Reasoning Up the Instruction Ladder for Controllable Language Models. In Findings of the Association for Computational Linguistics: ACL 2026, pages 39332-39354.
Download Paper