Fully open reproduction of DeepSeek-R1

Maya: An Instruction Finetuned Multilingual Multimodal Model using Aya

Grassroots Science Website

potato: portable text annotation tool

Processed / Cleaned Data for Paper Copilot

This repository contains datasets accompanying our paper Leveraging Large Language Models for Code-Mixed Data Augmentation in Sentiment Analysis (EMNLP '24). The datasets include both natural and s…

Githun Repo for “Preference Tuning with Human Feedback on Language, Speech, and Vision Tasks: A Survey”

EZswitch is a framework designed to generate code-switched text, blending two languages within a single sentence or discourse. This tool incorporates Equivalence Constraint Theory (ECT) with and (L…

WorldCuisines is an extensive multilingual and multicultural benchmark that spans 30 languages, covering a wide array of global cuisines. (NAACL 2025 Main Conference)

RewardBench: the first evaluation tool for reward models.

MetaMetrics is a calibrated meta-metric designed to evaluate generation tasks across different modalities aligned with alignment with human preferences.

A curated list of research papers and resources on Cultural LLM.

Resources for cultural NLP research

A Python module for getting the GPU status from NVIDA GPUs using nvidia-smi programmically in Python

[NeurIPS'23 Oral] Visual Instruction Tuning (LLaVA) built towards GPT-4V level capabilities and beyond.

ANSI color formatting for output in terminal

Mexican NLP 2024 Summerschool Tutorial on Knowledge Distillation and Parameter Efficient Finetuning

A Python implementation of global optimization with gaussian processes.

A library to calculate similarity scores between two collections of text sequences encoded using transformer models for bitext mining, dense retrieval, retrieval-based classification, and retrieval…

A library of translation-based text similarity measures

BERT score for text generation

MTEB: Massive Text Embedding Benchmark

Implementation of ProxyLM, a scalable and efficient LM performance prediction framework on NLP task using proxy models

Generate synthetic labeled data for extremely low-resource languages using bilingual lexicons.

MINERS ⛏️: The semantic retrieval benchmark for evaluating multilingual language models. (EMNLP 2024 Findings)

A collaborative project to collect datasets in SEA languages, SEA regions, or SEA cultures.