Modelpedia
Work in progress
About
Findings
Models
Concepts
Methods
Datasets
Sources
Light
Dark
Measuring and Enhancing Trustworthiness of LLMs in RAG through Grounded Attributions and Learning to Refuse
2025-01-22
· ICLR 2025 Oral ·
anchor
Findings
IC-170
GPT-4, GPT-3.5, and Claude-3.5-Sonnet rely heavily on parametric knowledge in RAG settings, producing ungrounded responses with high answered ratios and low trust-scores
IC-171
ICL prompting produces binary response patterns in released LLMs, with answered ratios collapsing to near 0% or 100% rather than calibrated refusal, making prompting ineffective for RAG groundedness