Improve the practical width/runtime frontier for fault-tolerant ECDLP attacks
on secp256k1.
Only edit:
ecdsa_autoresearch/candidate.py
Do not edit:
ecdsa_autoresearch/evaluator.pyecdsa_autoresearch/baselines.pyecdsa_autoresearch/toy_curves.pytests/
Comparable to 2017 and 2020 constructions.
Comparable to 2026-style low-space, multi-run constructions.
Reject the candidate if:
- correctness tests fail
- success probability is below the track target
- total Toffoli volume exceeds the track budget
- run count exceeds the track budget
- measurement model does not match the track configuration
Among valid candidates, compare lexicographically on:
(logical_qubits, toffoli_total, depth)
- dirty ancilla reuse
- register reuse
- pebbling
secp256k1-specific arithmetic- window-size tuning
- signed-digit representations
- RNS decomposition
- Legendre-style compression
- semi-classical / feedback-heavy variants
- phase-estimation changes
- curve-model changes with explicit conversion overhead
Do not mix track-changing candidates into the wrong leaderboard.