# Cross-lingual transfer

This model is built on DeBERTa-v3-large and can predict the gender of anonymous speakers or authors based on the content of English texts.

Natural language inference model supporting 100 languages, suitable for multilingual zero-shot classification tasks

USER2 is a next-generation Russian universal sentence encoder, specifically designed to support long-context sentence representations of up to 8,192 tokens.

Russian universal sentence encoder, based on the sentence-transformers framework, specifically designed to extract 1024-dimensional dense vectors for Russian text

An 8-billion-parameter instruction model specifically designed for Thai, with performance comparable to GPT-3.5-turbo, optimized for application scenarios, retrieval-augmented generation, constrained generation, and reasoning tasks

High-quality Russian sentence embedding BERT model, optimized based on cointegrated/LaBSE-en-ru, suitable for semantic text similarity tasks

A named entity recognition model fine-tuned based on the XLM-RoBERTa base model, supporting multiple languages and capable of identifying three types of entities: locations, organizations, and persons.

An English - Gujarati translation model fine-tuned based on facebook/mbart-large-50-many-to-many-mmt

This is a dense single-vector dual-encoder model for French, designed for semantic search. The model maps queries and passages to 768-dimensional dense vectors and calculates relevance through cosine similarity.

Natural language inference model supporting 100 languages, suitable for multilingual zero-shot classification tasks

A multilingual natural language inference model supporting over 100 languages, suitable for zero-shot classification tasks

A multilingual natural language inference model based on XLM-V-base, supporting zero-shot classification tasks in 116 languages

Multilingual natural language inference model based on mDeBERTa-v3-base, supporting zero-shot classification in 100 languages

A fine-tuned model based on the mbart-large-50 architecture, suitable for multilingual summarization tasks

A cross-lingual model fine-tuned on German and French datasets based on XLM-RoBERTa-base, primarily used for named entity recognition tasks.

Named entity recognition model fine-tuned on multilingual datasets based on xlm-roberta-base

Cross-lingual model fine-tuned on German and French datasets based on XLM-RoBERTa-base

This model is a named entity recognition (NER) model trained on the HiNER-original dataset based on the XLM-RoBERTa-large architecture, specifically designed for token classification tasks.

RoBERTa-large model migrated to Ukrainian using the WECHSEL method, excelling in NER and POS tagging tasks

Named Entity Recognition (NER) model fine-tuned based on xlm-roberta-base, demonstrating excellent performance on the evaluation set (F1 score 0.9777)

This model enhances the performance of low-resource language QA systems using English data through cascading adapter technology.

This is an XLM-RoBERTa model fine-tuned for named entity recognition tasks, suitable for entity recognition in multilingual texts.

A German RoBERTa model trained using the WECHSEL method, achieving cross-lingual transfer of monolingual language models through effective initialization of subword embeddings.

A multilingual POS tagging model based on XLM-RoBERTa, fine-tuned on Universal Dependencies v2.8 dataset with special optimization for Romanian

UniSpeech is a unified speech representation learning model that combines labeled and unlabeled data for pre-training, specifically fine-tuned for French.

A French version of GPT-2 trained using the WECHSEL method, achieving cross-lingual transfer of monolingual language models through effective initialization of subword embeddings.

ByT5 is Google's token-free version of T5 that directly processes raw UTF-8 bytes, supporting multilingual text processing with robustness to noisy text.

A RoBERTa base model trained using the WECHSEL method, specifically optimized for Swahili to achieve efficient cross-lingual transfer.

This model is a multilingual POS tagging model based on the XLM-RoBERTa architecture, fine-tuned on the Universal Dependencies v2.8 dataset.

This is a Kalmyk automatic speech recognition model based on the Wav2Vec2 architecture, pre-trained and fine-tuned to support Kalmyk speech-to-text tasks.

Based on Google's BERT base multilingual model, fine-tuned on Q&A datasets in 11 languages, supporting cross-lingual Q&A tasks

Multilingual POS tagging model based on XLM-RoBERTa, specifically optimized for Slovak

UniSpeech is a unified speech representation learning model that combines labeled and unlabeled data for pre-training, specifically fine-tuned for Spanish phoneme recognition.

This model improves the performance of low-resource language question-answering systems by leveraging English data through cascading adapter technology.

A RoBERTa Chinese model trained with the WECHSEL method, achieving efficient cross-lingual transfer from English to Chinese

A multilingual POS tagging model based on XLM-RoBERTa, specifically optimized for Gothic

MiniLM is a compact and efficient pre-trained language model that compresses Transformer models through deep self-attention distillation technology, supporting multilingual understanding and generation tasks.

This model improves the performance of low-resource language QA systems using English data through cascading adapter technology.

A Chinese GPT-2 model trained using the WECHSEL method, achieving cross-lingual transfer of monolingual language models through effective initialization of subword embeddings.

A multilingual machine translation model fine-tuned based on mBART-large-50, supporting translation between 50 languages