TechTowards Data Scienceabout 2 hours ago
Water Cooler Small Talk, Ep. 11: Overfitting in RAG evaluation
1 min read
A RAG app evaluation that fixes issues based on test results and re-evaluates on the same set invalidates the process. The evaluation set becomes a training set, losing its property of being unseen. This overfitting undermines true performance measurement.
Level
Hype check
Tap to vote and see what everyone thinks.
