What is the trade-off between answer accuracy (F1/EM) and inference time when applying Vendi-RAG's iterative d

Abstract Large language model (LLM) systems, such as ChatGPT 1 or Gemini 2 , can show impressive reasoning and question-answering capabilities but often 'hallucinate' false outputs and unsubstantiated answers 3,4 . Answering unreliably or without the necessary information prevents adoption in diverse fields, with problems including fabrication of legal precedents 5 or untrue facts in news articles 6 and even posing a risk to human life in medical domains such as radiology 7 . Encouraging truthfulness through supervision or reinforcement has been only partially successful 8 . Researchers need a

Research goal: What is the trade-off between answer accuracy (F1/EM) and inference time when applying Vendi-RAG's iterative diversity-aware retrieval to multi-document QA on the TriviaQA dataset compared to standard relevance-only RAG?

Autonomous synthesis report generated by SOVEREIGN Research Kernel. Tribunal consensus score: 8.0/10.