Skip to content

v8.2.1

Choose a tag to compare

@tatopenn-cell tatopenn-cell released this 24 Sep 05:07
· 310 commits to main since this release
8c5691e

New in this release:

  • coherence_predictive_zne_density_matrix added to dense_evolution.mitigation (#299): a coherence-informed ZNE signal covering phase-type noise (phaseflip, dephasing) the existing JSD-predictive signal is blind to by construction (it only reads the density-matrix diagonal). Validated at 200 seeds on phaseflip (63/200 active, 63/63 improve, p=1.07e-08), a noise-level sweep (significant through base_p<=0.10), and a second circuit family (VQE-style ansatz).
  • dense_evolution.protocols subpackage added (#300): BB84, three-party device-independent conference key agreement via GHZ(3) (Ribeiro, Murta & Wehner 2018), and the full multi-round DICKA structure, promoted from Dense-Evolution-Discovery's crypto-q RFC (issue #189) after passing that project's own promotion checklist.
  • Fixed order-dependent JAX x64 test flakiness via pytest-env (#301): jax_enable_x64 is enabled lazily by dense_evolution.config, which combined with pytest-randomly (already in CI) made precision order-dependent within a test session -- reproduced directly (11/55 failures in random order), fixed by forcing JAX_ENABLE_X64=1 before test collection.
  • Track SCF energy history and diagnose convergence via Dense-Armor (#291): HFResult.energy_history keeps the full per-iteration trace; new diagnose_convergence() in native_hf.scf runs Dense-Armor's Hampel/Tukey filters over it to flag oscillation instead of trusting only the final converged flag. New optional armor extra.
  • Add exact + FFT mass-decomposition utility (#292), promoted from CASMI26 spectral-identification work: parse_formula, rdbe, build_reachable_masses (exact subset-sum) and build_reachable_density_fft (FFT/convolution route), RDBE-filtered for chemical validity.
  • Add ia_utils.rag: hybrid TF-IDF/dense/cross-encoder retrieval (#293), promoted from the quantumrag literature tool. New optional rag extra.

PyPI: https://pypi.org/project/dense-evolution/8.2.1/