Advanced Retrieval Strategies
Properties
tags
gen_aigen_ai/rag
created
02.08.2026, 10:00
modified
02.08.2026, 10:10
published
Empty
sources
Empty
topics
RAG Retrieval, Vectorless Retrieval, Lexical Retrieval
authors
Opus 4.8
ai-assisted
Yes
- RAG methods whose innovation is the retrieval mechanism itself — how candidate documents are matched and fetched, often replacing or augmenting plain dense vector search.
# Topics
- Page Index
- Vectorless, reasoning-based retrieval: builds a table-of-contents tree and navigates it via LLM tree search, no vector DB or chunking.
- LATTICE - LLM-guided Hierarchical Retrieval (
Link)
- LLM navigates a semantic tree with logarithmic search complexity, using calibrated latent relevance scores for reasoning-intensive queries.
- GrepRAG
- Lightweight lexical retrieval where the LLM issues grep/ripgrep commands, refined with identifier-weighted BM25 reranking for code completion.
- RuleRAG
- Rule-guided retrieval that injects symbolic rules as in-context demonstrations to steer retrievers and generators.