There are several important SVP and CVP benchmarks in mathematics and computer science:
TU Darmstadt Lattice Challenge
https://www.latticechallenge.org/
The most well-known benchmark is the TU Darmstadt Lattice Challenge (available at latticechallenge.org), which provides lattice bases for testing and comparing lattice reduction algorithms. These challenges are designed so that solving SVP in one lattice implies solutions for all lattices of a certain smaller dimension, making them particularly suitable for worst-case analysis.
Recent Progress
The benchmark has seen steady progress over the years:
-
As of February 2021, researchers from CWI (Centrum Wiskunde & Informatica) solved the SVP Challenge for dimension 180 in 52 days using specialized hardware with 1.5 TB of RAM and four AI-capable graphics cards.
-
In June 2019, dimension 157 was solved using the General Sieve Kernel (G6K) implementation.
-
Earlier work solved a 128-dimensional ideal lattice challenge, which required approximately 30,000 CPU hours.
Darmstadt LWE Challenge
There's also the Darmstadt LWE (Learning With Errors) Challenge, which is related to lattice-based cryptography security. This challenge is particularly relevant for assessing the security of post-quantum cryptographic schemes.
Why These Benchmarks Matter
SVP and CVP are central to lattice-based cryptography, which is being considered for post-quantum cryptography since these problems are believed to be hard even for quantum computers. These benchmarks help researchers evaluate algorithm performance and inform secure parameter choices for next-generation cryptographic systems.