Peer review NPDMVchi

1. "Why no DiffusionCLIP?" DiffusionCLIP targets pixel-space diffusion guidance. Dynavect's contribution targets feed-forward latent-space modulator for StyleGAN. We nonetheless included two diffusion editors(SD-Img2Img which was published in 2023 & InstructPix2Pix) to bracket pixel-space performance. We will be happy to add a small DiffusionCLIP comparison on our public cohort in the revision. It will not affect our core claim about context aware latent deltas + identity projection. 2. Reduction of entanglement quantitatively We will include a targeted measurement: per-image delta(attribute) from a frozen classifier (e.g., gender for "blonde", age for "age+") and report mean delta(CI) across seeds. Our pilot analyses indicate reduced drift vs. StyleCLIP-G when identity orthogonalization is enabled.