AbridgeKV: Compressing Long Reasoning into the KV Cache of Learned Pause Tokens

(github.com) by matt_d | Oct 16, 2026 | 0 comments on HN