Context Dependence and Reliability in Autoregressive Language Models

Poushali Sengupta, Shashi Raj Pandey, Sabita Maharjan +1

Large language models (LLMs) generate outputs by utilizing extensive context, which often includes redundant information from prompts, retrieved passages, and interaction history.…