Cross-layer latent vector quantization enables extreme compression of embeddings from each layer.
OPTIC compresses Gemma 4 embeddings about 100 times with shared codes, retaining 97.8% bfloat16 performance and reducing GPU memory 29.7%.
OPTIC compresses Gemma 4 embeddings about 100 times with shared codes, retaining 97.8% bfloat16 performance and reducing GPU memory 29.7%.