Glossary · AI
What is
Context Rot?
A degradation in LLM quality as context length grows, even within the model's stated window.
By Anish· Founder · Vedwix
·Definition
Context rot is the observation that LLM accuracy decays as context length increases — even well within the stated context window. It manifests as forgotten earlier instructions, ignored retrieval results, or drift in agent behavior. Mitigations: smarter chunking, summarization, structured context, and avoiding context bloat.
Example
A model has a 200k context window, but at 150k tokens its instruction-following degrades noticeably.
How Vedwix uses Context Rot in client work
We design for context budgets ~30-50% of stated capacity for production reliability.
Building with Context Rot?
We ship this.
If you're building with Context Rot in production, we can help — from architecture review to full implementation.
Brief us