← 研踪 PaperTrailACL Anthology · ACL 2026:acl:2026.acl-srw.54 · 记录 #1
Probing Bias Formation in Medical LLMs through Activation Steering
来源日期（仅到月份） 2026-07
来源记录日期 2026-07
雷达观测 2026-11-01 21:44
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 environments, as models may prioritize contextual faithfulness to a biased prompt over their internal parametric medical knowledge. This study introduces a mechanistic approach to identify and mitigate these failures in MedGemma-27B, isolating hint integration circuits using Sparse Autoencoders and geometric manifold analysis. Our findings reveal that sycophantic bias is a highly distributed and polymorphic concept, with biased reasoning routed through shifting dimensions across transformer layers. We identify the optimal layer for intervention and demonstrate that cluster-conditioned dynamic steering tailored to the geometric subspace of the prompt outperforms static global interventions, though it reveals a fundamental tension between bias resilience and the retention of internal parametric knowledge. This work proposes a principled framework toward clinical AI systems that are more robust and aligned with expert medical logic, demonstrating the potential of cluster-conditioned geometric interventions while characterizing the inherent trade-offs in clinical knowledge retention.
论文发现出处
这些页面提供论文线索，不代表同行评审或质量认证。
观测到的版本
平台记录号代表本站观测到的元数据变化，不是作者修订号。来源日期仅到月份或年份时，按该月或该年首日存档。
记录 #1 · 2026-07 · 最新观测版本
Probing Bias Formation in Medical LLMs through Activation Steering
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 environments, as models may prioritize contextual faithfulness to a biased prompt over their internal parametric medical knowledge. This study introduces a mechanistic approach to identify and mitigate these failures in MedGemma-27B, isolating hint integration circuits using Sparse Autoencoders and geometric manifold analysis. Our findings reveal that sycophantic bias is a highly distributed and polymorphic concept, with biased reasoning routed through shifting dimensions across transformer layers. We identify the optimal layer for intervention and demonstrate that cluster-conditioned dynamic steering tailored to the geometric subspace of the prompt outperforms static global interventions, though it reveals a fundamental tension between bias resilience and the retention of internal parametric knowledge. This work proposes a principled framework toward clinical AI systems that are more robust and aligned with expert medical logic, demonstrating the potential of cluster-conditioned geometric interventions while characterizing the inherent trade-offs in clinical knowledge retention.
查看此版本解读 →查看此版本原文 ↗