
Cryptography engineer, working on Ethereum Lean client, prev: core dev @FantomFDN; contrib: @prylabs, @ErigonEth; mscs @FU_Berlin
onchain
Joined April 2017
- Tweets1.2K
- Following3K
- Followers476
- Likes5.8K
Andrei Ivasko retweeted
Zero-knowledge proofs for LLMs “make it harder for attackers to hijack a chat, or enable safety evaluators to verify that the model they’re testing is the same one being served to customers.” However, proving LLM inference remains expensive and requires substantial cryptographic engineering.
Our CCS 2026 paper introduces zkTransformer: we automatically compile Einsum expressions into an efficient proof protocol and make exponentiation cheaper to prove. Together, these techniques make GPT-2 proving over 10× faster than zkGPT in our multi-token evaluation.
The first challenge is translating LLM tensor operations into something a proof system can check. Manually constructing these protocols is tedious and easy to get wrong.
Our EinSumcheck starts from familiar Einsum expressions and automatically compiles them into a sumcheck instance, including the tricky cases where native translation can break the completeness of the protocol.
The second challenge is nonlinear operations such as softmax. Prior work exploits structure in range tables to speed up lookups, but exponentiation tables were believed to be unstructured and expensive to prove.
StructuredExp exposes structure in exponentiation table , enabling prover work that scales sublinearly with the table size.
By integrating the above techniques, zkTransformer brings GPT-2 proving down to approximately 1.7 seconds per token (over 112 generated tokens).
Code: github.com/uiuc-kang-lab/zkT…
Blog post: ddkang.substack.com/p/zktran…
Work w/ Bing-Jyue Chen and Lilia Tang
New Anthropic research: progress on provable inference for large language models.
Here, we introduce FlashProve. Building on established cryptographic techniques, FlashProve allows developers to verify whether a given response came from a given AI model. It’s substantially more efficient than prior open-source methods for doing so.
These proofs have a range of applications. They can, for instance, make it harder for attackers to hijack a chat, or enable safety evaluators to verify that the model they’re testing is the same one being served to customers.
For our full write-up, see here: anthropic.com/research/prova…
Further progress in provable inference will likely be an important part of improving trust in AI deployment. To that end, we're open-sourcing FlashProve’s code, here: github.com/anthropics/FlashP…
We’re grateful to our collaborators: Alessandro Chiesa, @GiacomoFenzi, @gwberg95, @miranda__christ, Rahul Dalal (rdalal.github.io) and Ben Mayer (@transmissions11).
On my way home from @EFDevcon, I experienced a very poor air quality in Delhi. Just wondering what causes this level of pollution, and why does not the government address this issue?
1
125
In April, Tim Beyne and Vincent Rijmen from COSIC taught a two-day course on linear cryptanalysis for Master's students at KU Leuven. We recorded it and are now sharing the lectures with everyone: 8 videos on YouTube.
#Cryptography
🎥 Watch it here: youtube.com/playlist?list=PL…
Andrei Ivasko retweeted
Happy to announce new work with @kobigurk and @ronrothblum . We prove unique decoding radius proximity gaps for expander codes with sufficient expansion.
Expanders codes are used in prover efficient SNARKs like Brakedown and Bolt.
Andrei Ivasko retweeted
Cambridge PQ Research 2026 playlist:
youtube.com/playlist?list=PL…
How could Ethereum’s $1M Proximity Prize help shrink post-quantum SNARKs?
Stronger guarantees beyond the Johnson bound lead to smaller proofs at the same security level.
@alexanderlee314 and I covered all math fundamentals to understand this topic.
Watch our @EFDevcon workshop youtube.com/watch?v=nCr8azKD…
Andrei Ivasko retweeted
[New] Sumcheck bit packing (Matteo Frigo and abhi shelat) ia.cr/2026/2770
Andrei Ivasko retweeted
Replying to @EFDevcon
@EFDevcon in Mumbai was great! Despite somewhat smaller attendence (expected), there was a lot of focus, good content and meetings.
Focus on Formal Verification this time. (a few dozen presentations overall).
My bet the next one will be Open Silicon and Open Hardware.
Andrei Ivasko retweeted
The strawmap set a goal of Lean Ethereum by 2029—we now have a fork-by-fork plan to get there.
H* will be the last fork without lean features. I* will ship the leanVM and lean transactions. J* will add lean attestations, DA sampling and PCS. By L*, Lean Ethereum will be complete
Tom Wambsgans answers my question about recursion challenges. It turned out, GKR allows efficient recursion for keccak in LeanVM . @leanEthereum @ethereum @EFDevcon
10
456
Andrei Ivasko retweeted
I'll be giving a talk today at 11 AM on Lotus Stage 1!
app.devcon.org/schedule/diam…
This talk isn't just for cryptographers!
In the first half, I'll discuss how obfuscation could enable entirely new Ethereum applications beyond finance, which is "Cryptoepistemics," in contrast to "Cryptoeconomics."
In the second half, I'll introduce Diamond iO 2!
Andrei Ivasko retweeted
New paper by @error024, Benedikt Wagner, and Emile Wambsgans on formally proving the security of hash-based signatures in Lean.
The goal: establish precise, machine-checked security bounds to help design better autoresearch challenges for future hash functions.
We published a paper (eprint.iacr.org/2026/2626) about exact definitions of bit-security. We noticed that depending on how you account the adversary's work we have slightly (but noticeable) different security of Hash-based schemes. Be careful with your Lean-proofs and statements :)
Andrei Ivasko retweeted
New Anthropic research: progress on provable inference for large language models.
Here, we introduce FlashProve. Building on established cryptographic techniques, FlashProve allows developers to verify whether a given response came from a given AI model. It’s substantially more efficient than prior open-source methods for doing so.
These proofs have a range of applications. They can, for instance, make it harder for attackers to hijack a chat, or enable safety evaluators to verify that the model they’re testing is the same one being served to customers.
For our full write-up, see here: anthropic.com/research/prova…
Further progress in provable inference will likely be an important part of improving trust in AI deployment. To that end, we're open-sourcing FlashProve’s code, here: github.com/anthropics/FlashP…
We’re grateful to our collaborators: Alessandro Chiesa, @GiacomoFenzi, @gwberg95, @miranda__christ, Rahul Dalal (rdalal.github.io) and Ben Mayer (@transmissions11).
I like @Nico_Mohnblatt explanation of the difference between post quantum and elliptic curve cryptography. Pq cryptography is a steel, elliptic curve cryptography is a wood. Ethereum will be made of steel! @EFDevcon
3
254
I was glad to meet old friends and meeting new folks at Speakers dinner at @EFDevcon. I am pretty sure, this will be a great conference in this vibrant city Mumbai.
3
9
386


















