Unfortunately, I think the context rot paper [1] found that the performance degr... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		mamp 8 months ago \| parent \| context \| favorite \| on: Claude Sonnet 4 now supports 1M tokens of context Unfortunately, I think the context rot paper [1] found that the performance degradation when context increased still occurred in models using attention sinks. 1. https://research.trychroma.com/context-rot

giancarlostoro 8 months ago [–]

Saw that paper have not had a chance to read it yet, are there other techniques that help then? I assume theres a few different ones used.

Consider applying for YC's Summer 2026 batch! Applications are open till May 4
Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact