Brief Details: KR-FinBert-SC is a Korean financial BERT model specialized in sentiment classification, achieving 96.3% accuracy on financial text analysis.
Brief-details: BERT-based model fine-tuned on GLUE MRPC dataset achieving 85.54% accuracy and 89.74% F1 score, optimized for text classification tasks.
Brief-details: Japanese Sentence-BERT model (v2) with 111M parameters, optimized with MultipleNegativesRankingLoss for better sentence embeddings. Built on BERT-base Japanese.
Brief-details: Japanese T5 model pre-trained on 100GB corpus (Wikipedia, OSCAR, CC-100). 222M parameters, outperforms mT5 on news classification with 97% accuracy.
Brief-details: A 1.2B parameter Korean language model by SK Telecom, trained on Ko-DAT dataset for text generation. Excels at Korean text tasks with strong reasoning capabilities.
Brief-details: A 66.4M parameter DistilBERT model fine-tuned for sentence similarity tasks, maps text to 768-dimensional vectors, developed by sentence-transformers
Brief Details: Powerful 4.86B parameter sentence embedding model based on T5, optimized for sentence similarity with 768-dimensional vectors, FP16 precision
Brief Details: A BERT-based dense passage retrieval model optimized for semantic search, offering 768-dimensional embeddings with 109M parameters and strong MS MARCO performance.
BRIEF DETAILS: A Facebook DPR context encoder model that maps sentences to 768-dimensional vectors, optimized for semantic search and clustering with 109M parameters.
Brief Details: Large-scale Russian BERT model (427M params) for sentence embeddings, optimized for NLU tasks with multi-task learning capabilities.
Brief-details: Russian variant of CLIP model combining ViT-B/32 image encoder with ruGPT3Small text encoder for text-image understanding, achieving 78% accuracy on CIFAR10
BRIEF DETAILS: Turkish BERT model fine-tuned for question-answering tasks using TQuAD dataset. 111M parameters, supports F32 tensors, ideal for Turkish NLP applications.
Brief-details: Turkish BERT-based sentiment analysis model with 95.4% accuracy, trained on 48K movie/product reviews and tweets, supporting binary classification
Brief-details: Turkish NER model based on BERT with 111M parameters. Achieves 92.5% F1 score on Turkish named entity recognition tasks. Trained on WikiAnn dataset.
Brief-details: DistilBERT-based industry classifier for categorizing business descriptions into 62 industry tags, trained on 7000 Indian company samples
Brief Details: Japanese GPT-2 small variant (123M params) trained on CC-100 and Wikipedia. MIT-licensed transformer model for Japanese text generation.
Brief Details: A domain-specific RoBERTa model fine-tuned on 4.6GB of legal corpora, optimized for legal text analysis with strong performance on legal MLM tasks.
Brief Details: A powerful sentence transformer model that maps text to 1024-dimensional vectors, optimized for semantic similarity tasks with BERT architecture.
Brief Details: GPT-2 model fine-tuned for Turkish language generation, trained on OSCAR corpus with 52K byte-level BPE vocab. Supports both PyTorch and TensorFlow.
BRIEF-DETAILS: Agriculture-focused BERT model trained on 6.5M paragraphs of agricultural texts, specialized for domain-specific NLP tasks and masked language modeling.
Brief Details: A BERT-based language model specialized for chemical industry texts, fine-tuned on 40,000+ technical documents with 250,000+ chemical domain tokens.
