Re:Making models smarter and cheaper at the same time
by Wetime on Monday July 20, 2026 @04:50AM
(#48975993)
Byte-exact KV grafting stores verified reasoning on disk. Replaying it lets a frozen 12B LLM beat 31B models at 8,700x less energy.
[ Parent
| on HN ]
There are no comments on this comment yet.