🤖 AI Summary
This work addresses a critical limitation in traditional Byzantine fault tolerance (BFT), which assumes honest nodes correctly enforce protocol semantics—an assumption that fails in agent-based systems where compliant nodes may erroneously endorse semantically invalid state transitions due to reasoning errors, thereby compromising execution safety. To resolve this, the paper introduces Epistemic Byzantine Fault Tolerance (EBFT), formally defining "epistemic faults" and the "honest majority problem," thereby decoupling semantic correctness from mere protocol compliance. The authors develop a two-dimensional fault-tolerance framework using confidence-weighted parameters: \(e_\delta\) for semantic safety risk and \(u_\varepsilon\) for liveness degradation. By integrating probabilistic belief modeling with tail concentration analysis, they derive novel quorum conditions that jointly guarantee semantic validity, consensus consistency, and system liveness. The analysis shows that fault tolerance improves only when newly added agents substantially reduce the tail risks of invalid endorsements or unavailable support.
📝 Abstract
State machine replication (SMR) and Byzantine fault-tolerant (BFT) consensus guarantee agreement despite a bounded number of arbitrary, colluding faulty participants. However, these guarantees rely on participants outside this set correctly executing the protocol's transition semantics. Agentic validators expose a weaker boundary: an authenticated, responsive, non-equivocating, and protocol-compliant reasoning participant may still endorse a semantically invalid transition due to reasoning errors.
We call this failure mode an epistemic fault, and the collective phenomenon the Honest Quorum Problem (where "honest" means protocol-compliant, not semantically correct). Such a quorum can satisfy ordinary checks while forming a certificate for an invalid transition. Thus, agreement alone does not guarantee semantic validity or execution safety. Furthermore, because agentic validators often share model weights, training distributions, prompts, or toolchains, they are highly susceptible to correlated epistemic faults.
We define Epistemic Byzantine Fault Tolerance (EBFT), a fault-tolerance model for agentic infrastructure and post-deterministic distributed systems. EBFT augments the conventional Byzantine fault bound with two separate, confidence-indexed quantities: $e_δ$ bounds coherent invalid endorsements outside the Byzantine set, and $u_ε$ bounds unusable validator support that degrades liveness. These quantities characterize semantic safety risk and liveness degradation independently. We derive quorum-threshold conditions for semantic validity, consensus agreement, liveness, and feasible threshold selection, and outline a calibration methodology for estimating these budgets. We show that adding nominally distinct agents improves fault tolerance only when it measurably reduces the upper-tail concentration of invalid endorsements or unusable support.