Confidence Calibration and Abstention Logic in Grounded Agents
Models withhold uncertainty when tasked to be helpful, even when they know better.
Priya Subramaniam
Section
Grounding & Hallucinations
6 stories in Grounding & Hallucinations.
Lost-in-the-Middle Problem and Re-Ranking for LLM Context Windows
Re-ranking retrieval results puts high-confidence chunks where models actually attend to them.
Marcus Tewolde
Hallucination Detection With Faithfulness Scoring in RAG Systems
Decompose RAG answers into claims to catch fabrications that end-to-end scores miss.
Marcus Tewolde
Knowledge Cutoff Gaps and Live Web Grounding for Time-Sensitive Queries
Models confidently hallucinate answers about facts beyond their training window.
Marcus Tewolde
Web Retrieval APIs for LLM Grounding vs SERP Snippet APIs
LLM agents need richer passages than search snippets provide to reason accurately and cite sources.
Marcus Tewolde
Inline Citation Patterns for Grounded Agent Responses
How AI agents hide citation failures behind the appearance of grounded responses.
Darian Kwok