2026.2.13

New Features

• Improve the performance of the automorphism submodule by using an
  optimized color refinement algorithm, doing more efficient pruning
  using existing automorphisms without changing the base, and by
  comparing graphs by their refinement history instead of their
  adjacency matrix. Additionally, graphs with disjoint components are
  handled more efficiently by determining if components are isomorphic
  and explicitly handling "swap" automorphisms between components.

• Add support for returning a mapping to the original graph labels, as
  well as vertex and edge orbits expressed in the original graph labels.

Bug Fixes

• Fix a bug where generators could be missed if coset representatives
  could belong to multiple left transversals, potentially undercounting
  the number of automorphisms for certain types of graphs.

2026.1.23

New Features

• Add an AutomorphismComposite that averages over graphically related
  embeddings using a common set of nodes and edges. This makes use of
  the new automorphism functionality in dwave-experimental. Add a simple
  example demonstrating how the AutomorphismComposite mitigates for
  control errors in a similar manner to SpinReversalTransformComposite.
  See #32.

• The embedding examples are expanded to include a case with detector
  and target only, and an example of distillation of a fully sourced and
  detected Chimera graph. A multicolor annealing example is presented
  demonstrating effectiveness of flux shimming for zero magnetization,
  and anneal-offsets for target qubit frequency synchronization.
  See #31.

Bug Fixes

• Update multicolor-annealing live tests to reflect the split of
  minTimeStep property into minAnnealingTimeStep and
  minPolarizingTimeStep. See #39.

2025.12.9

New Features

• Add dwave.experimental.automorphism submodule with a function to calculate
  the Schreier-Sims representation of a graph, as well as the number of
  automorphisms, vertex orbits, and edge orbits.

• Add dwave.experimental.multicolor_anneal submodule with a utility
  function to fetch available annealing lines and their properties.

• Add support for Python 3.14.

• Add support for solver name pattern matching in fast reverse anneal
  schedules. See #23.

• "x_polarizing_schedules" in sampling_params is detected as a
  symmetry-breaking parameter, allowing the symmetry-detection
  functionality in shimming routines to operate correctly for multicolor
  annealing.

  A utility is added that assigns anneal lines to Advantage2 processor
  nodes under the 6-annealing line scheme most often used. This function
  does not require a client, but is compatible with client-returned
  color assignments.

Upgrade Notes

• Drop support for Python 3.9.

Other Notes

• Add examples of embedding graphs with line-wise assignment
  restrictions.

2025.11.28

Other Notes

• Update solver name in fast-reverse-anneal schedules data to
  "Advantage2_research1.4". Schedules data remains unchanged.

2025.10.24

Other Notes

• Update solver name in fast-reverse-anneal schedules data to "Advantage2_research1.3". Schedules data remains unchanged.

2025.10.3

New Features

• Allow shimming to zero magnetization for a set of programmings differing by sampling_params.
  See #12.

• Implement hypergradient descent (arXiv:1703.04782B) for automation of learning rates as default option
  in shim_flux_biases(). See #10.

Bug Fixes

• Fix incorrect inversion of predicted fast-reverse-anneal schedules in time by updating the linear-exponential line shape parameters. See #14.

• Upgrade dependency packages version bounds to support Ocean 9.

2025.8.13

New Features

• Add a flux bias shimming utility function, dwave.experimental.shimming.shim_flux_biases. See #2.

• Add dwave.experimental.fast_reverse_anneal submodule with a utility function to fetch available annealing parameters and their extended type information (such as allowed values, etc). See #3.

• Add fast reverse anneal solver schedules and utility plotting functions. See #4.

• Setup dwave-experimental package and build infrastructure. See #1.

2025.8.13rc1

New Features

• Add a flux bias shimming utility function, dwave.experimental.shimming.shim_flux_biases. See #2.

• Add dwave.experimental.fast_reverse_anneal submodule with a utility function to fetch available annealing parameters and their extended type information (such as allowed values, etc). See #3.

• Add fast reverse anneal solver schedules and utility plotting functions. See #4.

• Setup dwave-experimental package and build infrastructure. See #1.