DictionaryContext Management
Context Rot
The longer a conversation gets, the worse the AI recalls details from early on, like a fading memory.
Definition
As more tokens are added to a conversation, the model’s ability to accurately retrieve specific pieces of information from the context decreases. Long conversations suffer from degraded attention, making early details harder to recall.
Example
Two hours into a session, the agent forgets the naming rule you gave in the first message and starts using camelCase again.
Appears in
- When and How Context Rot Appears in Coding Agents: A White-Box Study of Agent Skills in Code AuditingarXiv
- Diagnosing and Mitigating Context Rot in Long-horizon SearcharXiv
- Context Rot in AI-Assisted Software Development: Repurposing Documentation Consistency for AI Configuration ArtifactsarXiv
- Classifier Context Rot: Monitor Performance Degrades with Context LengtharXiv
- The $\mathbf{Y}$-Combinator for LLMs: Solving Long-Context Rot with $λ$-CalculusarXiv
- DISCO: Distributed Long Context Scaling with Grounding-Reasoning DisaggregationarXiv