Proof-of-Thought Secures Decentralized AI Coordination against Byzantine Malice
Proof-of-Thought, a novel consensus primitive, secures multi-agent LLM systems by rewarding the quality of reasoning, mitigating Byzantine collusion.
Asymmetric Quorums Enable Provable Subjective Trust in DAG Consensus
A new model for asynchronous consensus replaces the universal trust assumption with subjective node-specific quorums, enabling formally verifiable safety in flexible, open networks.
Hybrid Synchronous BFT Model Achieves Low Latency by Separating Message Sizes
AlterBFT introduces a hybrid synchronous model, relying on small, timely coordination messages to drastically reduce consensus latency.
Formal Sidechain Consensus Achieves Provable Safety and Liveness for Scaling
Introducing Cumulus, a formally proven sidechain consensus protocol that uses mainchain smart contracts to enforce block finality, ensuring scalable and secure interoperability.
Zero-Knowledge Proof of Training Secures Private Decentralized Machine Learning Consensus
Zero-Knowledge Proof of Training (ZKPoT) leverages zk-SNARKs to validate collaborative model performance privately, enabling scalable, secure decentralized AI.
Signature-Free Asynchronous Byzantine Agreement Achieves Optimal Communication Complexity
This new signature-free asynchronous Byzantine agreement protocol achieves the theoretical optimal communication complexity for unauthenticated consensus.
Zero-Knowledge Proof of Training Secures Private Decentralized Federated Consensus
ZKPoT is a new cryptographic primitive using zk-SNARKs to verify model contribution without revealing private data, unlocking decentralized AI.
Formalizing Accountable Finality Quantifies Proof-of-Stake Reorganization Economic Cost
The new Accountability Gadget formally quantifies the economic cost of PoS reorganizations, transforming finality from a social consensus into a provable, suicidal economic guarantee.
Zero-Knowledge Proof of Training Secures Private Decentralized Federated Learning
ZKPoT consensus verifiably proves model contribution quality via zk-SNARKs, fundamentally securing private, scalable decentralized AI.
