# Knowledge-intensive tasks

This project provides a quantized version of academic-ds-9B, aiming to make knowledge accessible to everyone.

Qwen3-30B-A3B-abliterated is a 30 billion parameter large language model based on the Qwen architecture, focusing on text generation tasks.

Llama-xLAM-2-70b-fc-r is a large language model released by Salesforce, based on the Llama 2 architecture with 70 billion parameters.

This is a 12B-parameter-scale language model merged using the mergekit tool, combining features from multiple excellent models based on EtherealAurora-12B-v2

ReasonIR-8B is the first retrieval model specifically trained for general reasoning tasks, achieving state-of-the-art retrieval performance on the BRIGHT benchmark and significantly improving performance on MMLU and GPQA benchmarks in RAG applications.

Llama-3.1-Tulu-3.1-8B is a large language model developed by AllenAI, based on the Llama architecture with 8 billion parameters, focusing on text generation tasks.

DeepSeek-R1-Distill-Llama-70B is a 70B parameter large language model developed by the DeepSeek team based on the Llama architecture. It is optimized through distillation technology and supports efficient inference and fine-tuning.

Turkish large language model based on Gemma-2-9b developed by WiroAI, specializing in dialogue generation tasks

Kyara is a language model fine-tuning project enhanced by knowledge retrieval, focusing on improving the model's performance on languages with fewer resources such as Traditional Chinese.

360 Zhinao is an open-source large language model series developed by Qihoo 360, including base models and dialogue models with various context lengths, supporting both Chinese and English.

StrangeMerges_53-7B-model_stock is the result of merging multiple 7B-parameter models using LazyMergekit, possessing powerful text generation capabilities.

A large language model fine-tuned with QLoRA on the Open-Platypus-Chat dataset based on Jamba-v0.1, supporting conversational tasks

Swallow-MX-8x7b-NVE-v0.1 is a Mixture of Experts model based on Mixtral-8x7B-Instruct-v0.1 with continued pretraining, primarily enhancing Japanese capabilities.

Pasta-Lake-7b is a 7B-parameter language model created by merging the Test157t/Pasta-PrimaMaid-7b and macadeliccc/WestLake-7B-v2-laser-truthy-dpo models.

OpenHermes 2 Mistral 7B is an advanced language model fine-tuned based on Mistral-7B, primarily trained using GPT-4 generated synthetic data, excelling in conversation and instruction-following tasks.

Sheared-LLaMA-1.3B is an efficient language model obtained through structured pruning and continual pre-training based on LLaMA-2-7B

An open-domain question answering model trained on the complete Natural Questions (NQ) dataset, capable of answering various factual questions

LinkBERT is an improved BERT model pre-trained on English Wikipedia and hyperlink information, enhancing performance on knowledge-intensive tasks by capturing cross-document associations.

Google's T5 model, specifically designed for closed-book QA tasks, achieving high performance on natural questions through multi-stage training.

T5-xxl SSM NQ is a closed-book QA model based on the T5 architecture, achieving knowledge storage and retrieval through pretraining and fine-tuning.