Goal
Independently verify the T13 publication evidence bundle from a clean clone. The bundle links 8 manuscript claims to 16 result artifacts with commands, limitations, and SHA-256 checksums.
This task is intentionally lightweight: a valid reproduction can confirm the current bundle or report a divergence. Both outcomes are useful.
Minimal check
git clone https://github.com/JVLegend/HyperMix.git
cd HyperMix
git checkout 9cbff27
python3 -I scripts/verify_evidence_manifest.py
Expected output:
Evidence bundle verified: 8 claims, 16 files.
No scientific dependencies, datasets, GPU, or training run are required for this minimal check.
Acceptance criteria
Please reply with:
- operating system and architecture;
- Python version;
- exact HyperMix commit;
- complete verifier output;
- whether any claim, path, byte count, or checksum diverged;
- if there was a divergence, the smallest reproducible description of it.
A successful verification and a failed verification are both accepted evidence. Please do not modify result files to make the check pass.
Optional scientific extension
If you have Python 3.11 and can install the training environment, also run:
python3.11 -m venv .venv-train
source .venv-train/bin/activate
pip install -e ".[train]"
pytest -q
Then report the dependency installation command, test count, failures, and total runtime.
Context
The scientific conclusion to preserve is that no learned method robustly beat the well-calibrated spatial matched filter in the completed regimes. This issue verifies traceability, not superiority.
Goal
Independently verify the T13 publication evidence bundle from a clean clone. The bundle links 8 manuscript claims to 16 result artifacts with commands, limitations, and SHA-256 checksums.
This task is intentionally lightweight: a valid reproduction can confirm the current bundle or report a divergence. Both outcomes are useful.
Minimal check
git clone https://github.com/JVLegend/HyperMix.git cd HyperMix git checkout 9cbff27 python3 -I scripts/verify_evidence_manifest.pyExpected output:
No scientific dependencies, datasets, GPU, or training run are required for this minimal check.
Acceptance criteria
Please reply with:
A successful verification and a failed verification are both accepted evidence. Please do not modify result files to make the check pass.
Optional scientific extension
If you have Python 3.11 and can install the training environment, also run:
Then report the dependency installation command, test count, failures, and total runtime.
Context
The scientific conclusion to preserve is that no learned method robustly beat the well-calibrated spatial matched filter in the completed regimes. This issue verifies traceability, not superiority.