一作ACL 2026Workshop· Natural Language Processing
Probing Bias Formation in Medical LLM s through Activation Steering
Bayram Ayadi · Annette Hautli-Janisz
摘要
Large Language Models specialized for the medical domain achieve high performance on static benchmarks, but remain vulnerable to sycophantic confabulation, where the models generate medically spurious rationales to justify incorrect user hints. This robustness gap poses severe risks in clinical e...