How an inference provider can prove they're not serving a quantized model

Posted by FrasiertheLion 18 hours ago

How an inference provider can prove they're not serving a quantized model(tinfoil.sh)

59 points | 44 commentspage 2

45dsilicon 4 hours ago|

[dead]

cmrx64 3 hours ago|

https://hellas.ai is building out their category theoretic compiler and protocol for solving this issue

tripplyons 2 hours ago|

ZKML is a very exciting emerging field, but the math is no where near efficient enough to prove an inference result for an LLM yet. They are probably just trying to sell their crypto token.