From Dev.to
LLM agents lose confidence under noisy retrieval
Even open‑weight retrieval agents keep shouting answers while the evidence they cite falls apart, exposing a systematic over‑confidence that survives obvious credibility warnings. The phenomenon only...
