00:17
2026-07-07
supercomputing-system-ai-lab.github.io
machine-learning
PuzzleMoE: Efficient Compression of Large Mixture-of-Experts Models
Researchers introduce PuzzleMoE, a method for compressing large Mixture-of-Experts models via fine-grained element-wise merging and bit-packing, achieving up to 16.7% higher accuracy on MMLU at 50% co…