---
id: "prereq-llm-context-windows"
type: "prereq"
source_timestamps: ["§2.2"]
tags: ["lineage", "context-window"]
related: ["claim-token-efficiency", "concept-context-scoping"]
sources: ["paper"]
sourceVaultSlug: "icm-paper-folder-architecture-2026Jun02"
originDay: 2
---
# LLM context windows and 'lost in the middle'

## Why this is a prerequisite

ICM's **efficiency argument** rests on keeping relevant tokens scoped and **out of the degraded middle** of a long context window.

## Key source

**Liu et al., "Lost in the Middle: How Language Models Use Long Contexts."**

Finding: models often perform *worst* on information placed in the middle of long inputs; careful placement and scoping of relevant content can help.

## What this supports

- The token-efficiency story in [[claim-token-efficiency]] (representative ~5k focused vs ~42k monolithic).
- The behavioural mechanism in [[concept-context-scoping]] (same model, different available info → different task).
- The layering design in [[concept-five-layer-hierarchy]] (relevance density via routing/content separation).

## Validation note

Liu et al.'s finding is **strong external support for the qualitative argument** (scoped contexts should outperform sprawling ones). It does **not** demonstrate the specific numerical advantage ICM cites — that requires a controlled comparison, the subject of [[question-controlled-comparison]].


## Related across days
- [[prereq-llm-context]]
- [[claim-token-efficiency]]
- [[concept-context-scoping]]
- [[concept-five-layer-hierarchy]]