Not All Heads Matter: A Head-Level KV Cache Compression Method with Integrated Retrieval and Reasoning · HackerPapers discussionshttps://hackerpaper.kaipfister.de/p/2410.19258/Collected HN comments and submissions linking this paper.enMon, 02 Nov 2026 11:16:22 GMTNot All Heads Matter: A Head-Level KV Cache Compression Method with Integrated Retrieval and Reasoning — mentioned by vlovich123https://news.ycombinator.com/item?id=42437226urn:hackerpapers:mention:2410.19258:42437226Tue, 17 Dec 2024 00:56:17 GMT<p>Wonder how this compares with Microsoft&#x27;s HeadKV paper [1] which claims a 98% percent reduction in memory while retaining 97% of the performance. [1] https://arxiv.org/html/2410.19258v3</p><p><a href="https://hackerpaper.kaipfister.de/p/2410.19258/">Paper and discussion history</a>. A mention is not an endorsement.</p>cs.CL