Confidence and Calibration of Activation Oracles for Reliable Interpretation of Language Model Internals 文章

ArXiv CS.CL2026-05-26NEWSen作者: Federico Torrielli, Peter Schneider-Kamp, Lukas Galke Poech

查看原文 →

Confidence and Calibration of Activation Oracles for Reliable Interpretation of Language Model Internals · 相关技术

相关技术

NAT ODE language model interpretability eval divide-and-conquer partitioning Token Superposition Training TPS Straight-Through Estimator SUL Prompt HTTP For CoT Camouflaged object detection Bootstrap ANN Git TPU