Validation set similarity increases leakage risk

Higher sequence similarity between validation and training sets increases the chance of data leakage, leading to overly optimistic performance estimates.

Part of Chain