New class of long-context tasks scales quadratically with corpus size, upending LCLM design choices

xiye_nlp · x · 2026-09-29

Researcher Prasann Singhal shares a thread asking whether long-context architectures can "find all contradictions" in science literature — not yet.

Original post →

More from Research

Research channel →