Query Expension for Better Query Embedding using LLMs
-
Updated
Feb 18, 2025 - Python
Query Expension for Better Query Embedding using LLMs
Code and models for the paper "Questions Are All You Need to Train a Dense Passage Retriever (TACL 2023)"
SPRINT Toolkit helps you evaluate diverse neural sparse models easily using a single click on any IR dataset.
Evaluation of BEIR Datasets using ColBERT retrieval model
LoRA fine-tuning of bi-encoder retrievers with hard negatives and cross-encoder distillation, evaluated on NFCorpus.
A genral RAG Search chatbot, with SoTA RAG techniques such as HyDE, Hybrid retrieval with BM25 + RRF and Cross encoder reranking. Evaluated on the BEIR scifact dataset and compared all the different pipelines i tried along the way
A decentralized cooperative memory & research layer for AI agents — collectively and cooperatively learning and advancing as a community.
Research-grade hybrid retrieval API — BM25 + FAISS + CDF calibration + entropy-weighted fusion + cross-encoder reranking. Benchmarked on BEIR SciFact with bootstrap significance tests.
Reproducible retrieval quality benchmark methods and curated result bundles for Cortrix semantic storage.
Retrieval benchmark that runs in CI without an API key: 4 strategies over BEIR corpora, scored with IR metrics, bootstrap intervals and paired significance tests
Rigorous retrieval evaluation over BEIR FiQA-2018: hand-written nDCG/recall/MRR/MAP validated against pytrec_eval to 1e-6, hybrid BM25+dense fusion, cross-encoder reranking, paired-bootstrap significance, and a single-use held-out test split.
Hybrid retrieval built to be measured, not believed — BM25 + dense + HNSW + fusion + re-ranking in pure numpy. Hand-rolled BM25 reproduces published BEIR baselines to 0.0066 MAD.
Retrieval evaluation that tells you what it can't resolve — intervals, significance, anytime-valid CI gates, and distribution-free guarantees. Zero dependencies, any retriever.
Retrieval bench: BM25, LSA and a dense bi-encoder behind one interface, evaluated on real BEIR benchmarks with nDCG/MRR and a per-query-type breakdown.
BM25 (Elasticsearch) vs dense (FAISS + MiniLM) vs hybrid RRF retrieval on BEIR SciFact, behind one Retriever protocol, scored on Recall, Precision, MRR and nDCG.
SERA-VQ: Discrete codes for extreme embedding compression — outperforms PCA+int8 at low memory budgets on BEIR/SciFact
Hybrid RAG pipeline combining BM25, dense retrieval, quantum kernels, Grover amplification and QAOA reranking.
RAG evaluation on BEIR SciFact: BM25, dense and hybrid retrieval with LLM answers.
端到端 Hybrid RAG 系统 (BEIR/SciFact):BM25 + Dense + Weighted Fusion + RAGAS 评估 + LLM-as-judge 稳定性研究
A RAG system that replaces standard BM25/FAISS retrieval with a fully learned neural retrieval stack - including a fine-tuned bi-encoder, a cross-encoder reranker, ColBERT-style late interaction scoring, and a locally hosted LLM generator. Built entirely with free and open-source tools.
To associate your repository with the beir topic, visit your repo's landing page and select "manage topics."