Adaptive Byzantine Consensus via Decentralized Reinforcement Learning
A reinforcement learning engine enables BFT protocols to dynamically self-optimize, boosting throughput and establishing the first Learned Consensus paradigm.
Deterministic Causal Structure Decouples Ledger Correctness from Ordering Policy
This theory introduces a Deterministic Causal Structure (DCS) where the ledger is a policy-agnostic DAG, resolving the entanglement of correctness and ordering.
Concurrent BFT Decouples Throughput and Latency, Eliminating Censorship
A new asynchronous BFT protocol concurrently executes dissemination and agreement, resolving the throughput-latency tradeoff and guaranteeing censorship resistance.
Optimality of BFT Responsiveness Achieves Minimal Network Latency
A new BFT lower bound proves the minimal latency trade-off, enabling consensus protocols to achieve theoretically optimal commitment speed in all network states.
