LostInTheMiddleRanker

Reorder documents so the most relevant ones appear at the beginning and end of the list, where LLMs tend to pay more attention.

Key Features

• Addresses the "lost in the middle" phenomenon where LLMs pay less attention to documents in the middle of a long context.
• Reorders documents so the most relevant ones appear at the beginning and end of the list.
• Works with any list of documents; no model inference required.
• Configurable number of documents to return via top_k.

Configuration

1. Drag the LostInTheMiddleRanker component onto the canvas from the Component Library.
2. Click on the component to open the configuration panel.
3. On the General tab:
   • Set top_k to control how many documents to return.

Connections

LostInTheMiddleRanker accepts a list of documents as input. Connect it after a retriever or DocumentJoiner in a query pipeline.

It outputs a reordered list of documents. Connect its documents output to ChatPromptBuilder or AnswerBuilder.

Source Code

To check this component's source code, open lost_in_the_middle.py in the Haystack repository.

Usage Examples

Basic Configuration

  LostInTheMiddleRanker:
    type: haystack.components.rankers.lost_in_the_middle.LostInTheMiddleRanker
    init_parameters: {}

Using the Component in a Pipeline

This query pipeline reorders retrieved documents before sending them to an LLM:

# haystack-pipeline
components:
  retriever:
    type: haystack_integrations.components.retrievers.opensearch.bm25_retriever.OpenSearchBM25Retriever
    init_parameters:
      document_store:
        type: haystack_integrations.document_stores.opensearch.document_store.OpenSearchDocumentStore
        init_parameters:
          index: default
      top_k: 20

  ranker:
    type: haystack.components.rankers.lost_in_the_middle.LostInTheMiddleRanker
    init_parameters:
      top_k: 10

  llm:
    type: haystack.components.generators.chat.llm.LLM
    init_parameters:
      chat_generator:
        type: haystack.components.generators.chat.openai.OpenAIChatGenerator
        init_parameters:
          model: gpt-4o-mini
          api_key:
            type: env_var
            env_vars:
            - OPENAI_API_KEY
            strict: false
      user_prompt: |-
        {% message role="user" %}
        Answer based on the documents.
        {% for doc in documents %}
        {{ doc.content }}
        {% endfor %}
        Question: {{ query }}
        {% endmessage %}
      required_variables:
      - query
      - documents

connections:
- sender: retriever.documents
  receiver: ranker.documents
- sender: ranker.documents
  receiver: llm.documents

inputs:
  query:
  - retriever.query
  - llm.query

outputs:
  messages: llm.last_message

max_runs_per_component: 100

metadata: {}

Parameters

Inputs

Parameter	Type	Default	Description
documents	List[Document]		A list of documents to reorder.
top_k	Optional[int]	None	The maximum number of documents to return.

Outputs

Parameter	Type	Default	Description
documents	List[Document]		Reordered list of documents with the most relevant ones at the beginning and end.

Init Parameters

These are the parameters you can configure in Pipeline Builder:

Parameter	Type	Default	Description
top_k	Optional[int]	None	The maximum number of documents to return.

Run Method Parameters

These are the parameters you can configure for the component's run() method. This means you can pass these parameters at query time through the API, in Playground, or when running a job. For details, see Modify Pipeline Parameters at Query Time.

Parameter	Type	Default	Description
documents	List[Document]		A list of documents to reorder.
top_k	Optional[int]	None	The maximum number of documents to return.