Skip to content

Latest commit

 

History

2 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 

Repository files navigation

File vs Vector Study Results

This repository contains the results and configuration of the five-domain retrieval study comparing keyword search and vector search performance. The results reflect derived MRR scores and indexing times without redistributing the original underlying textual passages (to adhere to license restrictions such as MS MARCO). For the full context and conclusions, read the study's blog post.

Datasets

  • CodeXGLUE
  • MS MARCO
  • SQuAD
  • HotpotQA
  • SciQ

Scale

  • 10k documents ingested per corpus.
  • 1k queries evaluated per corpus.

Scoring

  • Evaluated using MRR@10.

Hardware & Configuration

  • Keyword Search: Tantivy 0.22.0 default analyzer
  • Vector Search: ChromaDB ≥0.4.0 default index; all-MiniLM-L6-v2, 384-dim
  • Chunking: No chunking
  • Hardware: Apple M4, 16GB
  • Execution Date: Jan 14 2026

About

Results and configuration of a five-domain retrieval study comparing keyword (Tantivy) vs vector (ChromaDB) search performance.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors