POP: Prefill-Only Pruning for Efficient Large Model Inference Paper • 2602.03295 • Published Feb 3 • 4
SDAR Collection The models without suffixes use the default block size = 4. • 21 items • Updated Jan 2 • 9
WINA: Weight Informed Neuron Activation for Accelerating Large Language Model Inference Paper • 2505.19427 • Published May 26, 2025 • 11
