AnalysisAI Models
6 hours ago
Don't trust large context windows
Blog post argues that large context windows in LLMs are unreliable, citing issues with attention and accuracy over long inputs. Recommends not relying on extended context for critical tasks.