Analyze if 4-bit (P4) is the "Goldilocks zone" or if information loss in the vision encoder outweighs the memory savings.
is roughly 1/3 the size of base models; argue its viability for "Always-on" AI features.
How does the 4-bit quantization affect the embedding space compared to FP16?
Test on ImageNet-1K and CIFAR-100 .
Determine the "accuracy tax" paid for the extreme quantization. 2. Key Research Questions
Desired (short technical report vs. full journal paper)?
Analyze if 4-bit (P4) is the "Goldilocks zone" or if information loss in the vision encoder outweighs the memory savings.
is roughly 1/3 the size of base models; argue its viability for "Always-on" AI features.
How does the 4-bit quantization affect the embedding space compared to FP16?
Test on ImageNet-1K and CIFAR-100 .
Determine the "accuracy tax" paid for the extreme quantization. 2. Key Research Questions
Desired (short technical report vs. full journal paper)?