RESEARCHMonitorNEXT 12 MONTHS
LLMs Remember First, Forget Last: Dual-Process Interference in Large Language Models
arXiv cs.CL — Computation and Language
Factual evidence
What the source reports
Research shows LLMs experience dual-process interference and retrieval failures when handling conflicting or superseding contextual updates.
Open source