Skip to content

Add TAS-B, UniCOIL, SPLADE-v2, API embedding retrievers, PRP, TART, RankGemma, RankMistral#115

Merged
abdoelsayed2016 merged 5 commits into
mainfrom
copilot/add-retrievers-and-rerankers
Apr 23, 2026
Merged

Add TAS-B, UniCOIL, SPLADE-v2, API embedding retrievers, PRP, TART, RankGemma, RankMistral#115
abdoelsayed2016 merged 5 commits into
mainfrom
copilot/add-retrievers-and-rerankers

Conversation

Copy link
Copy Markdown
Contributor

Copilot AI commented Apr 23, 2026

  • Research 2025-2026 reranking papers with public code (castorini/rank_llm, April 2026)
  • MonoELECTRA — castorini/monoelectra-base pointwise ELECTRA cross-encoder
  • RankGemini — Google Gemini API listwise reranker (google-generativeai)
  • ReasonRank — CoT reasoning listwise reranker (<think>/<answer> format)
  • RankNemotron — NVIDIA Nemotron thinking model wrapper
  • RankQwen — Qwen/QwQ thinking model wrapper
  • All registered in METHOD_MAP and HF_PRE_DEFIND_MODELS
  • README: Cohere Rerank API marked ✅ (already implemented via APIRanker); new 5 models added
  • Fix spelling: normalise → normalize in api_embedding_retriever.py
  • Passed Code Review + CodeQL (0 alerts)

Not implementable (no public code found):

  • SetRank — no standalone 2025 public implementation on GitHub/HuggingFace
  • PolyEncoder — 2019 FAIR paper (ParlAI), dialogue-oriented; no 2025 reranking implementation
  • DynRank / ASRank — no public code repositories found

Copilot AI and others added 2 commits April 23, 2026 21:18
…/rerankers roadmap

Agent-Logs-Url: https://github.com/DataScienceUIBK/Rankify/sessions/2a8813be-c44c-4f3d-87d5-e22f3fea09e8

Co-authored-by: abdoelsayed2016 <27821589+abdoelsayed2016@users.noreply.github.com>
Agent-Logs-Url: https://github.com/DataScienceUIBK/Rankify/sessions/d738b02a-0683-4ab7-b0d2-6a6fce5ad791

Co-authored-by: abdoelsayed2016 <27821589+abdoelsayed2016@users.noreply.github.com>
Copilot AI and others added 2 commits April 23, 2026 21:41
…ankGemma, RankMistral

New retrievers:
- TAS-B: CLS-token DistilBERT dense retriever added to DiverDenseRetriever
- UniCOILRetriever: Pyserini LuceneImpactSearcher with UniCoilQueryEncoder
- SpladeV2Retriever: Pyserini LuceneImpactSearcher with SpladeQueryEncoder
- APIEmbeddingRetriever: OpenAI/Cohere/Voyage AI embedding + FAISS index

New rerankers:
- PRPReranker: Pairwise Ranking Prompting (allpairs + bubblesort modes)
- TARTReranker: Task-Aware Reranker with Instructions (cross-encoder)
- RankGemmaReranker: Gemma-based listwise reranker wrapping RankGPT
- RankMistralReranker: Mistral-based listwise reranker wrapping RankGPT

Registration:
- retriever.py METHOD_MAP + NO_INDEX_METHODS updated
- pre_defined_methods.py METHOD_MAP updated
- pre_defind_models.py HF_PRE_DEFIND_MODELS updated with aliases
- README.md roadmap and examples updated

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: abdoelsayed2016 <27821589+abdoelsayed2016@users.noreply.github.com>
- prp_reranker.py: Correct docstring to say O(N²) for bubblesort mode
- api_embedding_retriever.py: Add tqdm progress bar to _embed_openai

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: abdoelsayed2016 <27821589+abdoelsayed2016@users.noreply.github.com>
@abdoelsayed2016 abdoelsayed2016 marked this pull request as ready for review April 23, 2026 21:43
… rerankers; fix normalise→normalize

Agent-Logs-Url: https://github.com/DataScienceUIBK/Rankify/sessions/d5fac32f-63aa-46ab-b185-14a80881155f

Co-authored-by: abdoelsayed2016 <27821589+abdoelsayed2016@users.noreply.github.com>
@abdoelsayed2016 abdoelsayed2016 merged commit 5604499 into main Apr 23, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants