一作ICLR 2026Poster· Theory / Interpretability and Visualization
Hessian-Enhanced Token Attribution (HETA): Interpreting Autoregressive LLMs
Vishal Pramanik · Maisha Maliha · Nathaniel Bastian · Sumit Jha
首次获取该候选人的完整档案及 StartLux 评估共 20 积分；每位用户单独解锁，之后查看免费。主动刷新 20 积分。
实时查找电子邮箱，费用不超过20积分；未找到邮箱不扣费。
Vishal Pramanik · Maisha Maliha · Nathaniel Bastian · Sumit Jha
Vishal Pramanik · Maisha Maliha · Susmit Jha · Sumit Jha
Vishal Pramanik · Maisha Maliha · Susmit Jha · Alvaro Velasquez · Olivera Kotevska · Sumit Kumar Jha
Vishal Pramanik · Maisha Maliha · Nathaniel D. Bastian · Alvaro Velasquez · Susmit Jha · Sumit Kumar Jha
Chain-of-thought (CoT) prompting makes language models write step-by-step explanations, but these steps may not match what the model actually used to choose its answer. Existing faithfulness checks often only test whether changing the written chain changes the answer, without verifying whether th...