Lagrange Makes a Big Move: DeepProve-1 Enables LLM Inference to Self-Verify
Who would have thought that the zkML technology that could only be tinkered with on small models could now actually conquer LLMs?
@Lagrange Official The newly launched DeepProve-1 has directly become the first mass-production system capable of producing cryptographic proofs for complete large language model inference, and it successfully generated zero-knowledge proofs for the full inference of GPT-2.
What's even more impressive is that popular open-source models like GPT-2 and LLAMA are structurally very similar, which means that DeepProve is not far from supporting LLAMA as well.
In a few months, perhaps the open-source large models we commonly use will be able to use this verifiable feature.
