Research seriesL3algorithms
NIST standardized two very different signature philosophies. Comparing ML-DSA and SLH-DSA side by side, plus the stateful hash-based schemes, reveals the real trade-space: assumptions, size, speed, and the one operational rule that must never break. Grounded in FIPS 204, FIPS 205, SP 800-208, and the RFCs.
Before comparing post-quantum signature schemes, pin down what they must achieve: unforgeability against an adversary who can request signatures at will — and the two philosophies that reach it.
ML-DSA turns a lattice identification protocol into a signature, then rejection-samples every response so the transcript reveals nothing about the secret key.
SLH-DSA builds a stateless signature from nothing but a hash function, stacking one-time keys into Merkle trees and Merkle trees into a hypertree — trading large signatures for the most conservative assumption available.
XMSS and LMS give small, fast hash-based signatures by keeping a Merkle tree of one-time keys and a counter of which leaf is next — but reuse a single leaf and the scheme falls.
There is no universal winner among the post-quantum signatures. The right choice falls out of four axes — conservatism, size and speed, statefulness, and key lifetime — applied to your context.