Context Rot
A phenomenon observed during long-running AI agent tasks in which the context window becomes filled with logs, error messages, and intermediate results, causing the model's ability to effectively utilize key information to degrade over time.
Timeline (last 90 days)
"Context rot" refers to the phenomenon where many models nominally offer a 1M window but show a marked drop in recall in the middle of the sequence or under extremely long inputs.
The Cognition coding agent team described the phenomenon of 'context rot': models make weaker decisions at longer context lengths.
Too much irrelevant information, poor structure, or missing data can lead to worse reasoning and more hallucinations, a phenomenon known as 'Context Rot.'
All Facts (3)
"Context rot" refers to the phenomenon where many models nominally offer a 1M window but show a marked drop in recall in the middle of the sequence or under extremely long inputs.
50%UnverifiedThe Cognition coding agent team described the phenomenon of 'context rot': models make weaker decisions at longer context lengths.
50%UnverifiedToo much irrelevant information, poor structure, or missing data can lead to worse reasoning and more hallucinations, a phenomenon known as 'Context Rot.'
50%