ARTFEED — Contemporary Art Intelligence

Byzantine Agreement with Predictors: Resilience Trade-offs Characterized

other · 2026-05-20

A new paper on arXiv (2605.19452) presents a complete characterization of the trade-off between consistency and robustness in Byzantine Agreement when nodes have access to a predictor that flags potentially faulty nodes. The authors provide algorithms and impossibility results for both non-authenticated and authenticated settings. For n nodes and parameter α ∈ [0,1], their algorithms tolerate up to α·n faulty nodes when the predictor is correct (consistency) and up to (1-α)/2·n - 1 faulty nodes when the predictor is arbitrarily wrong (robustness). In the authenticated setting, the robustness bound improves to (1-α)·n - 1. These trade-offs are proven tight.

Key facts

  • Paper studies Byzantine Agreement with a predictor that flags suspicious nodes.
  • Focus is on algorithmic resilience depending on predictor accuracy.
  • Complete characterization of consistency–robustness trade-offs is provided.
  • Non-authenticated setting: consistency up to α·n, robustness up to (1-α)/2·n - 1.
  • Authenticated setting: robustness improves to (1-α)·n - 1.
  • Trade-offs are shown to be exactly tight.
  • Paper is on arXiv with ID 2605.19452.
  • Authors present both algorithms and impossibility results.

Entities

Institutions

  • arXiv

Sources