The prime repository for state-of-the-art Multilingual Question Answering research and development.

An easy to use Neural Search Engine. Index latent vectors along with JSON metadata and do efficient k-NN search.

An end-to-end neural ad-hoc ranking pipeline.

SWIM-IR is a Synthetic Wikipedia-based Multilingual Information Retrieval training set with 28 million query-passage pairs spanning 33 languages, generated using PaLM 2 and summarize-then-ask prompting.

Advanced IR Experiments with SBERT, Elasticsearch and trec_eval

A Configuration-Driven Framework for Reproducible Information Retrieval Experiments

Python code to train SPLADE sparse retrieval models based on BERT-Tiny (4M) and BERT-Mini (11M) by distilling a Cross-Encoder on the MSMARCO dataset

Achademic search engine exploiting Neural Information Retrieval approaches

Adaptive hybrid retrieval: per-query BM25 + dense fusion via QDAP (Query-Driven Alpha Prediction). PyTorch implementation with antagonist negative sampling, evaluated on MLDR & MIRACL (29 languages).