RepIt: Steering Language Models with Concept-Specific Refusal Vectors
This work addresses the need for more granular control over LLM behavior, particularly in counteracting overgeneralization, though it is incremental in refining activation steering methods.
The authors tackled the problem of isolating concept-specific representations in large language models to enable targeted interventions, achieving precise suppression of refusal on targeted concepts while maintaining safety on standard benchmarks with minimal data and compute requirements.
While activation steering in large language models (LLMs) is a growing area of research, methods can often incur broader effects than desired. This motivates isolation of purer concept vectors to enable targeted interventions and understand LLM behavior at a more granular level. We present RepIt, a simple and data-efficient framework for isolating concept-specific representations. Across five frontier LLMs, RepIt enables precise interventions: it selectively suppresses refusal on targeted concepts while preserving refusal elsewhere, producing models that answer WMD-related questions while still scoring as safe on standard benchmarks. We further show that the corrective signal localizes to just 100-200 neurons and that robust target representations can be extracted from as few as a dozen examples on a single A6000. This efficiency raises a dual concern: manipulations can be performed with modest compute and data to extend to underrepresented data-scarce topics while evading existing benchmarks. By disentangling refusal vectors with RepIt, this work demonstrates that targeted interventions can counteract overgeneralization, laying the foundation for more granular control of model behavior.