"Dataset for \"UQ-RAG: A Universal and Quality-Aware Framework for Adaptive Retrieval-Augmented Generation\""
Abstract
"While Adaptive Retrieval-Augmented Generation (RAG) has shown effectiveness in mitigating hallucinations, existing paradigms suffer from limited domain generalization and uni-dimensional evidence evaluation. In practice, many methods overfit to source domains, leading to severe calibra tion errors when transferred to unseen vertical domains, and remain vulnerable to evidence that is semantically relevant yet logically inconsistent. To address these challenges, we propose UQ-RAG, a unified framework for universal domain adaptation and fine-grained evidence quality perception. First, we introduce an Adversarial Transfer Learning mechanism with a Gradient Reversal Layer (GRL) to explicitly disentangle domain-specific surface features from query-level meta-features (complexity and intent). This design enables robust zero-shot generalization for retrieval decision-making in unseen domains. Second, we propose a knowledge-type-aware retrieval routing mechanism. Based on the query classification results, the model automatically deter mines whether external retrieval is required and further selects between domain-specific knowledge bases and open web search, enabling appropriate knowledge acquisition for different types of queries. Third, to move beyond purely semantic relevance, we develop a multi-dimensional quality distillation strategy. We distill the teacher large language model\u2019s ability to evaluate retrieved evidence along three dimensions\u2014relevance, logical…
Read the paper
Related studies
- Knowledge Management Perspective of Generative Artificial Intelligence
- Navigating Generative Artificial Intelligence Promises and Perils for Knowledge and Creative Work
- Characterizing the evolving research on enterprise content management
- Large language models and multimodal foundation models for precision oncology
- From Generative AI to Extended Reality: Multidisciplinary Perspectives on the Challenges, Opportunities, and Future of Educational Computing
- Generative Artificial Intelligence: A New Engine for Advancing Environmental Science and Engineering
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.