MLReproMutate v0.1.4

Documentation-focused release expanding Python API usage examples for software peer review.

MLReproMutate v0.1.3

Full Changelog: v0.1.2...v0.1.3

MLReproMutate v0.1.2

Patch release fixing sandbox import isolation for editable-installed projects.

Fixed:

• Sandbox validation now prioritizes the sandbox project source tree over inherited Python paths.
• Prevents false SURVIVED outcomes when an editable-installed original project would otherwise be imported outside the sandbox.
• Adds regression coverage for the reported scenario.

See #1 and #2.

MLReproMutate v0.1.1

MLReproMutate v0.1.1 improves CLI usability and deployment workflows.

Highlights:

• interactive CLI setup;
• static mutation candidate detection;
• sandbox path exclusions;
• in-place execution for CI/disposable workspaces;
• portable python/python3 command resolution;
• explicit mutation-operator selection;
• expanded user and CI documentation.

MLReproMutate v0.1.0

First public research-software release of MLReproMutate.

MLReproMutate provides mutation testing for reproducibility-relevant
safeguards in machine-learning research software.

This release includes:

• random-seed mutation;
• dependency-pin mutation with resolved-version semantic verification;
• data-split stratification mutation;
• cross-validation fold-count mutation;
• baseline-first mutation evaluation;
• machine-readable mutation reports;
• empirical-study accounting and provenance artifacts.

The empirical corpus and primary study outcomes are frozen and are not
modified by this software release.