{"id":"course-correction-steerability","title":"A Course Correction in Steerability Evaluation: Revealing Miscalibration and Side Effects in LLMs","authors":[{"given":"Trenton","family":"Chang"},{"given":"Tobias","family":"Schnabel"},{"given":"Adith","family":"Swaminathan"},{"given":"Jenna","family":"Wiens"}],"first_public":"2025-05-27","date_note":"First arXiv version.","date_source":"https://arxiv.org/abs/2505.23816","year":2026,"venue":"AAAI 2026","paper_url":"https://ojs.aaai.org/index.php/AAAI/article/view/41057","doi":"10.1609/aaai.v40i44.41057","arxiv":"2505.23816","category":"evaluation","relationship":"uses","relationship_label":"Uses WildVis","dataset_version":"English-language examples retrieved through WildVis validate the study’s text-rewriting goals; the controlled experiments use separate source-text corpora.","summary":"Uses keyword searches and manual review in WildVis to select realistic goals for text rewriting. The study tests whether models can change one property, such as reading difficulty or formality, while controlling other properties.","finding":"WildVis examples support the choice of rewriting goals. The experiments find miscalibration and unintended changes to other text properties.","evidence_url":"https://arxiv.org/html/2505.23816v1","evidence_note":"Appendix D.5: WildVis searches, English filtering, manual review, and conversation examples.","bibtex_key":"chang2026coursecorrection","record_url":"https://yuntiandeng.com/wildchat/papers/course-correction-steerability/","citation_title":"A Course Correction in Steerability Evaluation: Revealing Miscalibration and Side Effects in LLMs","citation_type":"paper-conference","bibtex_type":"inproceedings","container_title":"Proceedings of the AAAI Conference on Artificial Intelligence","citation_url":"https://doi.org/10.1609/aaai.v40i44.41057","citation_year":2026,"citation_note":"AAAI 2026 proceedings paper.","volume":"40","issue":"44","pages":"37259–37267","publisher":"AAAI Press"}