Fix champs-scalar-coupling to include test molecule structures

[jvpoulos]

The prepare script explicitly filters structures to only include train molecules. This is incorrect for this competition --- in the Kaggle competition, structures.csv contains all molecules (both train and test). The test molecules need their structures to make predictions, but they're being filtered out.

Changes:

• Modified prepare.py to include both train and test molecules in structures.csv
• Updated checksums.yaml to reflect the new structures.csv checksum
• Updated assertions to validate both train and test molecules

After fixing the data preparation issue, I used a LightGBM model trained on 25% of the data to make predictions:

  {
    "competition_id": "champs-scalar-coupling",
    "score": 0.5823,
    "gold_threshold": -2.87509,
    "silver_threshold": -2.03119,
    "bronze_threshold": -1.90122,
    "median_threshold": -0.9529,
    "any_medal": false,
    "gold_medal": false,
    "silver_medal": false,
    "bronze_medal": false,
    "above_median": false,
    "submission_exists": true,
    "valid_submission": true,
    "is_lower_better": true,
    "created_at": "2025-09-03T22:56:19.592098",
    "submission_path": "mlebench/competitions/champs-scalar-coupling/submission.csv"
  }

[thesofakillers]

Thank you for catching this. You are correct and there is a mistake in the prepare.py, and your fix seems right.

As explained in the readme in #66 we won't be merging this fix in yet, and will release it as a batch of fixes in a upcoming v2 to be released on openai/preparedness. I've added as tracked in #71. I will try to put you as co-author for when we release the fix.

For submissions to the v1 leaderboard, please proceed as if this issue was not present.