Get Started
Home
Topics
Search
Library
Research questionHow can models calibrate factual uncertainty to decide when external retrieval is needed?Language models may answer factual questions confidently even when their answers are wrong. Retrieving evidence for every question can improve reliability but also adds cost and latency when retrieval is unnecessary.
AI
Alignment & Safety
Evaluation & Benchmarks
Information Retrieval
LLM Pretraining & Post-training
Machine Learning
Natural Language Processing
Retrieval-Augmented Generation
Latest papersRecent research connected to this question, newest first.ConfRAG: Confidence-Guided Retrieval-Augmenting GenerationThe source fine-tunes language models on atomic factual statements to produce an uncertainty response and uses that response to trigger retrieval. Evidence is limited to factual question answering with external retrieval and the reported benchmark settings; it does not establish performance across all models, domains, or deployment conditions.research paper · Sep 4, 2026
Related questions
How can search agents learn when retrieval is necessary and ground answers in evidence without costly supervision?How can NER systems decide when external retrieval helps long-tail entities without adding noise or cost?How can retrieval-augmented language models resist ordinary-looking GEO-optimized documents that distort synthesized answers?How can retrieval-augmented generation reason causally over dynamic external information without losing context?