Lost in the Middle
A long-context failure mode where mid-document information is underused compared with beginning or end segments
#Lost in the Middle#context loss#mid-context degradation#long context
What is Lost in the Middle?
Lost in the Middle is a long-context behavior where models rely less on information in the middle of a document than on the beginning or end.
How does it show up?
As context length grows, models may skip central evidence and cite less relevant sections.
- Missing key evidence from middle sections
- Citation bias toward opening or closing segments
- Weaker reasoning across multi-part documents
Why does it matter?
This directly affects reliability in RAG, long-form QA, and report generation.
Mitigations typically include chunk design, reranking, and explicit context refresh.
Related terms
Natural Language Processing
AGI (Artificial General Intelligence)
A hypothetical AI system capable of performing any intellectual task a human can
Natural Language Processing
AI Agent
An autonomous AI system that can plan, use tools, and take actions to achieve goals
Natural Language Processing
Attention
A mechanism that allows AI models to focus on the most relevant parts of the input when producing output
Natural Language Processing
Chunk
A text segment created by splitting long documents into meaningful units for retrieval and generation
Natural Language Processing
Claude Opus
Claude's top-tier model family optimized for deep multi-step reasoning and high-stakes analysis
Natural Language Processing
Claude Sonnet
Claude's practical model family optimized for speed, cost efficiency, and strong day-to-day quality