1 paper
Kenji Sahay, Snigdha Pandya, Rohan Nagale +4
Large language models (LLMs) often generate fluent but factually incorrect statements despite having access to relevant evidence, a failure mode rooted in how they allocate attenti…