This was no easy task 🥵 We had to split GPT2-XL's 48 decoder blocks into 144 sub-blocks, before proving on our massive 1TB memory, 128 core CPU cloud clusters for nearly 90 hours! That means that — using generic prover systems — zkLLM overhead is over 1,000,000x!!!

𝐄𝐭𝐡𝐞𝐫𝐞𝐮𝐦 𝐡𝐚𝐬 𝐬𝐩𝐨𝐤𝐞𝐧 𝐢𝐭𝐬 𝐟𝐢𝐫𝐬𝐭 𝐰𝐨𝐫𝐝! That's right — we have completed the world's 1st ZKP of a 1B+ param LLM inference, before verifying the proof on Ethereum mainnet (block 19427725) 🤯 What did Ethereum say? Learn more in 🧵 medium.com/@ModulusLabs/chap…

Want to learn more? Dive into all the details with our paper, “Scaling Intelligence: Verifiable Decision Forest Inference with 𝑅𝑒𝑚𝑎𝑖𝑛𝑑𝑒𝑟” Pretty graphs and so much more await you 😉📈 at github.com/Modulus-Labs/Pape…

Our third key insight is generalizing the linear-time prover algorithm from Libra [XZZ+19] to the dataparallel case, allowing 𝑅𝑒𝑚𝑎𝑖𝑛𝑑𝑒𝑟 to achieve a strictly linear proving time for even dataparallel, unstructured layers requiring wiring predicates ♎-🦒

Our second key insight is a novel layered claim combination strategy, built over the interpolation protocol in Time-Optimal Interactive Proofs for Circuit Evaluation [Tha13] and augmented with a claim grouping heuristic, shaving proof generation time by ~40-60% ⌛⬇️

We're excited to unveil 𝑅𝑒𝑚𝑎𝑖𝑛𝑑𝑒𝑟—the world's most powerful ZKML prover Our first implementation achieved a mere 180x proof generation overhead vs. the same (non-verifiable) computation on an M2 Mac. Highlights below, read all about it here: medium.com/@ModulusLabs/chap… 🚀

Welcome to the family, RoboNet! Check out AI-powered DeFi today😍

AI security in crypto isn't just a goal, it's a journey 🗺️ With ✓Explorer, we're blending the best of indexers & dashes to help all of us become savvier AI consumers, continuing the ZKML security story! Our Stanford talk for those digging deeper: youtube.com/watch?v=lC2PD3kW…

Transaction page is where the real magic happens though 🔮 (1) Model Card: the “Ingredients”, with the model’s architecture, performance, historic authorship, and more 🥗 (2) Inference Card: the nitty-gritty details of the specific inference verified on-chain + zk metadata ☎️

It starts with the landing page: a realtime headline of [# of AI proofs] settled to Ethereum and [$ impacted] as a consequence 💸 And to the right, our partner roster ( @UpshotHQ @polychainmon @ionprotocol) and recent transactions + Search for specific verified AI inferences 📝

Initially an internal tool, the Modulus ✓Explorer slowly evolved into our favorite utility to explain ZKML to partners and clients 🤯 That's how it earned the moniker: the "Nutrition Facts" of ZK verified AI results (plus it works on mobile 📱)

⛵ Say hello to the Modulus ✓Explorer ⛵ We're tackling the ZKML consumer story with our new AI verification dashboard — the "Nutrition Facts" of ✓verified AI This human-friendly interface to the world of ZKML proving is live rn! Learn more: medium.com/@ModulusLabs/chap…