KV-cache compression

KVCapsule

KVCapsule: Efficient Sequential KV Cache Compression for Vision-Language Models with Asymmetric Redundancy

On the frontierfirst seen May 14, 2026

Current frontier — recent, not yet superseded in the knowledge base

0 papers critique it · 0 beat it on benchmarks

What to use instead

Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.