# Low-resource NLP

Statically quantized version based on boltuix/NeuroBERT-Mini, optimized for edge devices

SimpleStories is a micro model family designed for interpretability research, trained on the SimpleStories dataset, focusing on story generation tasks.

A few-shot summarization generation model based on BART-large, trained with 100 samples from the XSUM dataset, demonstrating the potential of few-shot learning in summarization tasks.

A cross-token large language model developed for the Tatar language, converted from Mistral-7B-Instruct-v0.2

A Tigrinya abusive content detection model fine-tuned on TiRoBERTa, trained on 13,717 YouTube comments dataset

A website classification model based on DistilBERT, achieving an accuracy of 95.04% on an unknown dataset through fine-tuning.

AfroLM is a pretrained language model optimized for 23 African languages, employing an active learning framework to achieve high performance with minimal data

This model is a fine-tuned version of BanglaBERT on the Bengali SQuAD dataset for question answering tasks

MyanBERTa is a Burmese pre-trained language model based on the BERT architecture, pre-trained on a Burmese dataset containing 5,992,299 sentences.

RoBERTa series models pretrained on datasets of varying scales (1M-1B tokens), including BASE and MED-SMALL specifications

This model is a text classification model fine-tuned on the GLUE RTE task based on ALBERT-large-v2, used for recognizing textual entailment relationships.

This is a small RoBERTa model based on Indonesian language, specifically optimized for Indonesian text processing tasks.

A RoBERTa base model pre-trained on 1B tokens with a validation perplexity of 3.93, suitable for English text processing tasks.

RoBERTa variants pre-trained on datasets ranging from 1M to 1B tokens, including BASE and MED-SMALL specifications, suitable for natural language processing tasks in resource-limited scenarios

ELECTRA is an efficient self-supervised language representation learning method that replaces traditional generative pretraining with discriminative pretraining, significantly improving computational efficiency.

IndicBART is a multilingual sequence-to-sequence pre-trained model focused on Indian languages and English, supporting 11 Indian languages, built on the mBART architecture.