This paper presents a novel construction for post-quantum key exchange protocols leveraging structured lattice problems. We demonstrate that our scheme achieves forward secrecy under the Learning With Errors assumption while maintaining practical key sizes and computational overhead comparable to existing elliptic curve methods.
We introduce a leaderless consensus protocol that achieves safety and liveness in fully asynchronous networks with up to one-third Byzantine nodes, requiring only O(n log n) message complexity per decision round.
We propose a hybrid architecture combining sparse attention mechanisms with state-space models for sequences exceeding 1M tokens, achieving linear scaling in both memory and computation while preserving in-context learning capabilities.
This paper formalizes a mechanism for funding digital public goods through voluntary contribution games with commitment devices. We prove that our mechanism is strategy-proof, individually rational, and achieves near-optimal social welfare under mild assumptions about contributor valuations.
We present a framework for generating zero-knowledge proofs over fully homomorphic encrypted data, enabling verifiable computation without decryption. Our construction reduces proof generation time by 3 orders of magnitude compared to existing FHE-ZKP combinations.
We redesign the state channel paradigm to support arbitrary smart contract execution with only a single on-chain transaction for channel opening and closing, achieving constant gas costs regardless of off-chain interaction complexity.