Patching

2000년 도입 · 논문 102편에서 사용

Activation patching studies the model's computation by altering its latent representations, the token embeddings in transformer-based language models, during the inference process

출처: Patchscopes: A Unifying Framework for Inspecting Hidden Representations of Language Models

소개 논문: Patchscopes: A Unifying Framework for Inspecting Hidden Representations of Language Models