liveBook · Manning

RA-DIT

As explained in Retrieval Augmented Generation, The Foundational Ideas

RA-DIT, or Retrieval-Augmented Dual Instruction Tuning, is a retrieval-augmented instruction-tuning method for training model components to cooperate.

RA-DIT (Retrieval-Augmented Dual Instruction Tuning) retrofits cooperation between an off-the-shelf retriever and language model through two lightweight fine-tuning passes, avoiding Atlas-style joint pre-training. section 6.6 section 6.6.1

The method separates the two training jobs:

  • Fine-tune only the query encoder with LM-Supervised Retrieval. A KL-divergence loss matches retriever scores to a distribution derived from each chunk’s improvement to the language model’s probability of the correct output. section 6.6.1
  • Freeze the document encoder so the existing index remains usable and never needs re-embedding. section 6.6.1
  • Fine-tune the language model on instruction examples with retrieved chunks prepended as background, using the ordinary next-token objective. section 6.6.1
  • Teach the reader to rely on retrieved passages rather than treating them as irrelevant prompt text. section 6.6.1

A miniature reader pass

Six training questions receive contexts where a distractor sentence precedes the answer passage. Fifteen optimization steps move two held-out answers from log-probabilities of -3.2 and -0.4 to approximately 0.0 and 0.0, showing the reader learning to use the supplied context. section 6.6.1

RA-DIT assumes access to both components for fine-tuning. When the language model is closed behind an API, REPLUG preserves the likelihood-based retrieval supervision but trains only the retriever against the frozen model. section 6.6.2

How Retrieval Augmented Generation, The Foundational Ideas covers it

How deep it goes
Explains both training passes, the KL signal, encoder freezing, and a miniature reader loop, but stops short of joint pre-training.
Examples
Uses six distractor-context training examples, two held-out questions, and a fifteen-step language-model fine-tuning loop.
What it assumes
Assumes an existing retriever, language model, indexed corpus, and instruction data, while teaching their cooperation through retrieved contexts.
What you can do after
Apply two lightweight fine-tuning passes to an existing retriever and language model while preserving the document index.
How it connects
Connects retrieval augmentation with instruction tuning, language-model likelihood, KL divergence, query encoders, document encoders, and REPLUG.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Retrieval Augmented Generation, The Foundational Ideas. The section links point to the parts of the book it draws on.