Compare learning curves against existing models on the same dataset

[forklady42]

Overview

Benchmark ElectrAI's learning curves against existing models trained on the same dataset to contextualize performance.

Tasks

• Identify existing baseline models to compare against (e.g., prior super-resolution or charge density models)
• Ensure all models are evaluated on identical train/val/test splits
• Collect or reproduce learning curves (loss vs. epoch, and/or loss vs. data size) for baselines
• Plot ElectrAI learning curves alongside baselines using consistent metrics
• Document findings — where does ElectrAI outperform or underperform baselines, and at what data/compute scales?

Acceptance Criteria

• At least one baseline model identified and compared
• Side-by-side learning curve plot produced
• Written summary of comparison included in results