MonoT5

Stub. Created as a placeholder — expand with vault-sourced content.

Definition

MonoT5 is a T5-based pointwise neural reranker: it frames relevance as a seq2seq task, scoring a (query, document) pair by the model’s probability of generating a “true” vs. “false” token.

Vault references (existing coverage)

TODO

  • Add the mono/duo distinction (DuoT5 = pairwise) and cite the source paper.