Selective Capability Unlearning in End-to-End Spoken Language Understanding
For developers of SLU systems needing to remove specific functionalities due to policy or safety constraints, this work addresses a structural failure in existing unlearning methods.
The paper identifies 'capability persistence' in SLU models where suppressing an intent does not remove its slot-generation behavior, and proposes BSU to attenuate this mapping, substantially reducing forced-prefix recoverability while preserving retained performance.
Modern spoken language understanding (SLU) systems are increasingly deployed in real-world settings, where specific functionalities may need to be removed due to policy or safety constraints. In SLU, a functionality corresponds to an intent and its associated slot-generation behavior. However, in autoregressive models, suppressing a target intent does not eliminate the conditional mapping that generates slots conditioned on that intent. When the intent prefix is externally supplied, the model can reconstruct the original intent-slot structure. We identify this structural failure as \textbf{\emph{capability persistence}}. We propose \textit{\underline{B}inding \underline{S}ubspace (BSU)}, a representation-level framework that isolates and attenuates intent-conditioned directions underlying this mapping. Across SLU benchmarks, BSU substantially reduces forced-prefix recoverability while preserving retained performance.