Applying OpenAI's RAG Strategies

we expand on each method mention and show how you can implement each one for yourself.

Baseline

The base-case retrieval method used in the OpenAI study mentioned cosine similarity.

distance metricsへの参考

Query Transformations (1)

Query transformations are a set of approaches focused on modifying the user input in order to improve retrieval.

Query expansion

OpenAIによって報告されなかったメソッド

Step back prompting

Rewrite-Retrieve-Read

Routing (2)

The OpenAI presentation reported that they needed to route question between two vectorstores and single SQL database.

SQLデータベースはtool利用の話ではないかと思われる

Query Construction (3)

valid SQL needed to be generated from the user input in order to extract the necessary information.

OpenAIによって報告されなかったメソッド

Text-to-metadata filter for vectorstores

Text-to-Cypher for graph databases

Text-to-SQL+semantic for semi-structured data in Postgres with Pgvector

Building the Index (4)

OpenAI reported an notable boost in performance simply from experimenting with the chunk size during document embedding.

報告されなかったメソッドに「embedding fine-tuning」

HuggingFaceのtutorial

Post-Processing (5)

We can use post-processing to enforce diversity or recency among our retrieved documents, which can be especially important when we are pooling documents from multiple sources.

Re-rank

RAG-fusion

Classification

This marries two ideas

tagging of text

logical routing

OpenAIによって報告されなかったメソッド

MMR

クラスタリング