@misc{indiciae11244d7f030a, title = {Controlling the Risk of Corrupted Contexts for Language Models via Early-Exiting}, author = {Andrea Wynn and Metod Jazbec and Charith Peris and Rinat Khaziev and Anqi Liu and Daniel Khashabi and Eric Nalisnick}, year = {2026}, url = {https://arxiv.org/abs/2510.02480}, note = {Source identifier: 2510.02480} }