Lost in the Middle
Definition
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.
Is your site visible in AI search?
See for free how ChatGPT, Perplexity, and Gemini describe your brand.
Start Free Diagnosis →Related terms
Natural Language Processing
Agentic AI
A category of AI systems that autonomously decompose goals, use tools, and run multi-step tasks
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
BigLaw Bench
A benchmark for legal-task performance, focusing on document interpretation and reasoning consistency
Natural Language Processing
Chain-of-Thought Elicitation
A prompting method that asks a model to reveal intermediate reasoning steps before the final answer