Grouped Value Attention: Efficient KV Caching via On-Demand Key Reconstruction Paper • 2609.13285 • Published 8 days ago • 60
Grouped Query Experts: Mixture-of-Experts on GQA Self-Attention Paper • 2606.20945 • Published Jun 18 • 81
ZClip: Adaptive Spike Mitigation for LLM Pre-Training Paper • 2504.02507 • Published Apr 3, 2025 • 90