Deep Funding Score Visualization

This interactive visualization demonstrates how changes in repository weight distributions affect the pairwise cost score in the Deep Funding competition.

Learn more here.

Using the Visualization

• Load Baseline: Upload comparison data (CSV with repo_a, repo_b, choice, multiplier columns) to generate optimized baseline weights, or upload pre-computed weights. This is the same format as used in the "train.csv" file in the competition. The default baseline is based on "train.csv" and synthetic data to ensure all repos were represented in the data.
• Load Submission: Upload alternative weight distributions to compare against the baseline. This is in the same format as submissions in the contest. The default submission was selected to show interesting differences.
• Interpolation Slider: Blend between baseline and alternative distributions to see how rankings change. When the slider is at zero, both sides are identical (i.e., both sides are at baseline). The slider interpolates from baseline to your submission. So, given an arbitrary baseline, you can see how many changes get you to a particular "score" - it's an interesting way to understand how far you might be from a perfect score.
• Manual Editing: Click on individual weights in the submission column to adjust them directly

The Pairwise Cost Score shows how well the submission weights match the baseline preferences from human comparisons - lower scores indicate better alignment.