SearchAbout
#ai-model-optimizationStory

KVpop -- Key-Value Cache Compression with Predictive Online Pruning

KVpop is a novel key-value cache compression method that uses predictive online pruning to achieve high performance while reducing memory usage. It learns a fixed-budget KV eviction policy by directly supervising the keep-or-drop decision. KVpop outperforms established eviction baselines on AIME and HMMT mathematical reasoning benchmarks.

#ai-model-optimization#autoregressive-decoding#key-value-cache-compression#memory-efficiency#research-method
Hugging Face Daily Papers3 min read1d ago
KVpop -- Key-Value Cache Compression with Predictive Online Pruning
Read original ↗Sign in to upvoteSign in to save
ShareXLinkedInHacker News

Related stories

Teaching models to forget: Selective unlearning with Amazon NovaAWS Machine Learning Blog12h agodOPSD: On-Policy Self-Distillation for Diffusion Language ModelsHugging Face Daily Papers2d agoGetting Claude Code to grunt in Caveman-speak might not save as many tokens as you thinkThe New Stack (AI)16h agoBreaking Failure Cascades: Step-Aware Reinforcement Learning for Medical Multimodal ReasoningHugging Face Daily Papers7d ago