Conceptrepresentation engineering / Representation Engineering
Representation Engineering
A systematic AI research framework formally introduced by Zou et al. at Stanford University in 2023. Its core insight is that high-level semantics in large language models can be edited and controlled through linear operations on internal representations, situated within the fields of AI transparency and interpretability.
Timeline (last 90 days)
Oct 28
Recent research on representation engineering and activation steering has shown that certain semantic or stylistic attributes (such as sentiment倾向 and refusal behavior) have approximately linear directions in activation space, and adding or subtracting vectors along that direction can modulate model output.
Unverified50%