179 questions
No questions match those filters.
← All questions
EvaluationHard· Recall· Graph Traversal· Error Analysis
Your rule-based extractor reports 95% precision, but ha...
This is one of the questions in the full AI/ML interview bank. Pro unlocks all 1789 questions; Premium includes the same bank plus the highest daily Practice limit.
See plansIf the graph is missing the path, no amount of reranking or generator improvement will fix the multi-hop failures. You should measure recall against hand-labeled sentences or an LLM-extracted ‘silver’ set to determine if the edges needed for traversal are actually present in the graph.