Now, the most powerful component is the Comparative Analysis tab. After selecting the best retriever based on previous insights, you can post a question here. The system then runs all three explainability scoring modes in parallel — Perturbation Dissimilarity, Retrieval Weighted, and Confidence Ret…
Now, the most powerful component is the Comparative Analysis tab. After selecting the best retriever based on previous insights, you can post a question here. The system then runs all three explainability scoring modes in parallel — Perturbation Dissimilarity, Retrieval Weighted, and Confidence Retrieval Fusion — according to the selected explanation level and perturbation strategy.
Here, we choose the sentence level with a leave-one-out perturbation strategy, so each unit corresponds to a sent