mirror of
https://github.com/lux-apps/papers.git
synced 2026-10-30 18:15:21 +00:00
zeekay
abf99d05fc
Restructured from flat .tex/.pdf files into per-paper subdirectories. Added Antje Worring as co-author on all papers. Synced PDFs to pdfs/ for download site.
27 lines
1.4 KiB
TeX
27 lines
1.4 KiB
TeX
\section{Introduction}
|
|
|
|
\label{sec:introduction}
|
|
|
|
Existing metastable consensus protocols use fixed parameters $(k, \alpha, \beta)$ tuned
|
|
for worst-case adversarial conditions. This means that even in benign network conditions
|
|
(no Byzantine validators, low latency), transactions must wait for the same number of
|
|
confirmation rounds as under attack. This paper introduces the Ray protocol, which
|
|
adapts its finality threshold in real-time based on observed network behavior.
|
|
|
|
The key insight is that the \emph{rate} at which confidence accumulates carries
|
|
information about the adversarial environment. Rapid, unanimous convergence suggests
|
|
benign conditions where fewer confirmation rounds suffice. Slow, contested convergence
|
|
suggests adversarial interference requiring more rounds.
|
|
|
|
\paragraph{Contributions.}
|
|
\begin{enumerate}[nosep]
|
|
\item The Ray adaptive finality protocol with confidence velocity metric (\S\ref{sec:protocol}).
|
|
\item An online Bayesian adversary estimator derived from query response distributions (\S\ref{sec:estimator}).
|
|
\item A proof that adaptive thresholding preserves $2^{-128}$ safety (\S\ref{sec:safety}).
|
|
\item A 40\% latency reduction compared to static protocols in benign conditions (\S\ref{sec:evaluation}).
|
|
\end{enumerate}
|
|
|
|
%% -----------------------------------------------------------------------
|
|
%% 2. SYSTEM MODEL
|
|
%% -----------------------------------------------------------------------
|