Gemini: Elastic SNARKs for Diverse Environments
summary
Gemini claims a single SNARK construction that admits either a logarithmic-space prover or a linear-time prover while producing the same verifier-facing proof; it targets R1CS with streaming access, implements a streaming inner-product folding algorithm (n log n time, log space) and switches to an in-memory algorithm once a memory budget is reached. Implementation in arkworks uses KZG commitments, multi-scale multiplication and Ruffini division for evaluation proofs, and demonstrates proofs of up to 2^37 gates using ~500 MB memory.
AI-written summary · about
chapters (8)
- 0:00Succinct arguments motivation
- 3:14Streaming SNARKs elasticity
- 7:49Elastic inner product design
- 12:43Polynomial commitments
- 16:39Implementation challenges
- 18:05Benchmarking performance
- 20:10Memory budget considerations
- 24:33Conclusion Q&A
related talks
23mSonic: Nearly Trustless SNARK Setups
39mPrivacy and Evolution of SNARKs
52mFaster, Smaller, Weaker, Darker
31mSecure Distributed Storage
18mA Selection of Pairing Based zk-SNARKs
19mDIZK: A Distributed Zero Knowledge Proof System
01h 09mProofs - Practice 1
27mRecursive SNARKs for Efficiency, Scalability, and Privacy