Submissions will be evaluated using Mean Squared Error (MSE) between the predicted compatibility scores and the ground-truth compatibility values provided in `target.csv`.

Each row corresponds to a pair of participants, and the MSE is computed over all valid (`src_user_id`, `dst_user_id`) pairs. Lower MSE indicates better performance.

The evaluation encourages accurate and consistent estimation of pairwise compatibility scores across the entire participant set.

## Submission format

Participants must submit a CSV file containing pairwise compatibility predictions with the following columns:
```csv
src_user_id+'_'+dst_user_id, compatibility_score
```

## Format Details

- **`src_user_id:`** Profile ID of the source participant

- **`dst_user_id:`** Profile ID of the destination participant

- **`compatibility_score:`** Predicted compatibility score between the two participants

## Submission Constraints

- The submission must have same number of rows as the possible number of pairs in test data i.e. 16,000
- Ensure that all the entries are numbers and **headers are saved in csv file.**
- All the compatibility scores must be between 0 and 1.
- Ensure the ID row is string

## Sample submission
```csv
ID, compatibility_score
5601_5601,1
5601_5602,0.1538
5601_5603,0
5601_5604,0.2
```

## Leaderboard

* The **Public Leaderboard** reflects scores computed on the public portion of the validation data during the competition.
* Final rankings will be determined using a **Private Leaderboard**, evaluated on a hidden validation set.
* In case of a tie on the final leaderboard, the team with the **earlier valid submission** will be ranked higher.