Local-GenAI-Search is a generative search engine based on Llama 3, langchain and qdrant that answers questions based on your local files

CauseNet: Towards a Causality Graph Extracted from the Web

Machine Comprehension Train on MSMARCO with S-NET Extraction Modification

Analysis on the MS-MARCO leaderboard regarding the machine reading comprehension task.

This repo shows how foundational model hallucinates and how we can fix such hallucinations using fine-tuning them

Fixed some bugs and edited original baseline scripts from MS MARCO

阅读/论文笔记

MONSERRATE is a dataset specifically created to evaluate Question Generation systems. It has, on average, 26 questions associated to each source sentence, attempting to be an “exhaustive” reference.

A Fine-Grained Analysis of Distribution Shifts in MSMARCO (MS-Shift). Evaluation benchmark on three types of distribution shifts, all conditioned on MSMARCO queries.

Information Retrieval pipeline leveraging TFIDF and BM25 to evaluate ranking accuracy over MSMARCO Passage dataset

Final project for course on deep learning for nlp (IA376E/1s2020 @ Unicamp)

Information Retrieval Using TFIDF and Conv-KNRM

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

Doc-Phi: An Intelligent Document Finder

Sample dataset for test JUDGO

Fine-tuning of transformer-based models to predict semantically relevant outlinks in large-scale web crawling. Includes binary classification, score estimation with QualT5, and a multimodal BERT architecture integrating textual and metadata features for robust frontier prioritisation.

Learning-to-Rank on MS MARCO Passages: candidate generation from prebuilt indexes and re-ranking for QA search

In this demo, we illustrate the the possibility of using Semantic Search + Recognising Textual Entailment with Gradio to build an automated fact checking tool

Document search engine built for the MSMARCO collection. Multimedia Infomation Retrieval @ University of Pisa

AniSearchModel leverages Sentence-BERT (SBERT) models to generate embeddings for synopses, enabling the calculation of semantic similarities between descriptions. This allows users to find the most similar anime or manga based on a given description.