# Multi-task Adaptation

DeepSeek-R1-Distill-Qwen-14B is an optimized large language model with a parameter scale of 14B, released by DeepSeek AI. It is distilled from the Qwen architecture and offers multiple GGUF quantization versions to improve performance.

This is a fine-tuned sentence-transformers model specifically optimized for financial document retrieval tasks while maintaining general performance.

Hugging Face's Transformer model library supporting various natural language processing tasks

The fifth-generation Qwen2.5-YOYO model integrates features from multiple advanced models, optimizes the model merging formula, and supports a context length of 1 million tokens.

Cuckoo is a small (300M parameters) information extraction model that efficiently extracts information by mimicking the next-word prediction paradigm of large language models

No specific model information provided, likely a generic transformers library model

Doge 20M is a small language model based on dynamic masked attention mechanism, supporting instruction following and Q&A tasks.

Ruri is a Japanese universal text embedding model, focusing on sentence similarity calculation and feature extraction, with support for long text processing.

Conan-embedding-v1 is a Chinese text embedding model developed by the Tencent BAC team, implemented based on the sentence-transformers library, suitable for various Chinese natural language processing tasks.

This is a sentence-transformers model fine-tuned from Alibaba NLP/gte-multilingual-base, specifically designed for Persian semantic text tasks.

A Turkish BERT model pre-trained by Trendyol, suitable for various natural language understanding tasks.

xFinder-llama38it is a fine-tuned key answer extraction model based on Llama3-8B-Instruct, designed to improve the accuracy and robustness of key answer extraction from large language model outputs.

The world's first 6-billion-parameter Llama-3 base model, created using the downgrade loop technique from Meta-Llama-3-8B and continuously pretrained on 1 billion English text tokens

InBedder is a text embedder specifically designed to follow instructions, capable of capturing text features specified by user instructions through question-answering.

MAEST is a series of Transformer models based on PASST, focusing on music analysis applications, capable of classifying 400 music genres

Instructor is an instruction-tuned text embedding model capable of generating customized text embedding vectors for any task and domain without additional fine-tuning.

This is a plain old model with unspecified specific functions and purposes.

OPT is an open-source pre-trained Transformer language model developed by Meta AI, with parameter scales ranging from 125 million to 175 billion, designed to advance research in large-scale language models.

This is a Chinese pretrained BERT model using whole word masking strategy, released by the Harbin Institute of Technology-iFLYTEK Joint Laboratory to accelerate Chinese natural language processing research.

KoELECTRA is a Korean pre-trained language model based on the ELECTRA architecture, optimized for Korean natural language processing tasks.

GPT-fr is a French GPT model developed by Quantmetry and the Laboratoire de Linguistique Formelle (LLF), trained on a large and diverse French corpus.

Large-scale biomedical language model based on BERT, ALBERT, and ELECTRA, specialized for biomedical domain tasks

A pre-trained unbalanced Transformer model for Chinese understanding and generation, supporting various natural language processing tasks

DeBERTa V2 XXLarge is an enhanced natural language understanding model developed by Microsoft, which improves the BERT architecture through a disentangled attention mechanism and enhanced masked decoder, achieving SOTA performance on multiple NLP tasks.

Large-scale biomedical language models based on BERT, ALBERT, and ELECTRA, achieving state-of-the-art results in multiple biomedical tasks

Chinese ELECTRA is a Chinese pre-trained model based on Google's ELECTRA, released by the HIT & iFLYTEK Joint Lab (HFL), featuring a compact structure and superior performance.

This is the first Hindi language model trained based on Google's ELECTRA framework, suitable for various Hindi NLP tasks.

This is a sparsely pretrained BERT-Base model achieving 90% weight sparsity through one-shot pruning, suitable for fine-tuning on various language tasks.

This model is fine-tuned from bert-base-uncased-sparse-70-unstructured on the MNLI task (GLUE benchmark), with the classifier layer removed for easier loading into other downstream tasks for training.