Semantic Visual Reconstruction

Add explicit visual supervision to VLMs by training models to autoregressively reconstruct semantic image tokens, achieving 2-3% average gains and 10-point improvements on hallucination robustness.

adu2021 Updated 7 repo stars

File contents

adu2021/skillxiv commit a2ac37f07c

Frequently asked questions

npx get-skill add adu2021/semantic-visual-reconstruction