TempPerturb-Eval: On the Joint Effects of Internal Temperature and External Perturbations in RAG Robustness

ArXi:2512.01183v2 Announce Type: replace-cross The evaluation of Retrieval-Augmented Generation (RAG) systems typically examines retrieval quality and generation parameters like temperature in isolation, overlooking their interaction. This work presents a systematic investigation of how text perturbations (simulating noisy retrieval) interact with temperature settings across multiple LLM runs. We propose a comprehensive RAG Perturbation-Temperature Analysis Framework that subjects retrieved documents to three distinct perturbation types across varying temperature settings.