liveBook · Manning

MGFiD

As explained in Retrieval Augmented Generation, The Foundational Ideas

MGFiD is a Fusion-in-Decoder architecture that selects fine-grained evidence and represents key sentences with an anchor vector.

MGFiD (Multi-Granularity Guided Fusion-in-Decoder) focuses answer generation on fine-grained evidence instead of passing every retrieved passage to the decoder equally. section 5.6.1

Its processing pipeline is:

  • Re-rank the retrieved passages. section 5.6.1
  • Select the most important sentence from each top-ranked passage. section 5.6.1
  • Aggregate those sentences into an anchor vector. section 5.6.1
  • Prune less important passages and provide the decoder with the anchor vector plus the remaining context. section 5.6.1

This multi-granularity design improves exact match on benchmarks while reducing decoder load by concentrating reads on finer-grained evidence. It extends the FiD progression beyond passage-level fusion toward selective, more interpretable context. section 5.6.1

How Retrieval Augmented Generation, The Foundational Ideas covers it

How deep it goes
Defines reranking, sentence selection, anchor aggregation, pruning, and decoding, without detailing implementation internals.
Examples
Uses top-ranked passages, selected key sentences, an anchor vector, and exact-match benchmarks as concrete artifacts.
What it assumes
Assumes familiarity with Fusion-in-Decoder, passage retrieval, and decoder attention.
What you can do after
Evaluate/understand how sentence-level evidence selection focuses Fusion-in-Decoder context and reduces decoder load.
How it connects
Connects Fusion-in-Decoder with retrieval, reranking, decoder attention, evidence granularity, and context pruning.

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.