4 ms·
This is well established at this point, it’s called “context rot”: https://research.trychroma.com/context-rot https://research.trychroma.com/context-rot
by taejavu 1y ago
This is well established at this point, it’s called “context rot”: https://research.trychroma.com/context-rot https://research.trychroma.com/context-rot
- cubefox 1y agoYeah, though this paper doesn't test any standard LLM benchmarks like GPQA diamond, SimpleQA, AIME 25, LiveCodeBench v5, etc. So it remains hard to tell how much intelligence is lost when the context is filled with irrelevant information.