m-a-p/Qwen2.5-Instruct-7B-COIG-P

TEXT GENERATIONConcurrency Cost:1Model Size:7.6BQuant:FP8Ctx Length:32kPublished:Apr 2, 2025License:cc-by-nc-4.0Architecture:Transformer Open Weights Cold

The m-a-p/Qwen2.5-Instruct-7B-COIG-P is a 7.6 billion parameter Large Language Model, fine-tuned from Qwen2, specifically designed for instruction following in Chinese. It leverages the COIG-P dataset, which includes 101k Chinese preference pairs across diverse domains like Chat, Code, Math, and Logic. This model excels at text generation tasks and is particularly well-suited for Chinese language processing, offering capabilities for creative text, translation, and question answering.

Loading preview...

Overview

The m-a-p/Qwen2.5-Instruct-7B-COIG-P is a 7.6 billion parameter Large Language Model (LLM) that has been fine-tuned from the Qwen2 base model. Its primary focus is instruction following, particularly within the Chinese language context. The model's development utilized the COIG-P dataset, a high-quality, large-scale Chinese preference dataset comprising 101k pairs across six domains: Chat, Code, Math, Logic, Novel, and Role. This specialized training aims to align the model with human values and preferences in Chinese.

Key Capabilities

  • Chinese Language Processing: Optimized for generating and understanding text in Chinese.
  • Instruction Following: Designed to accurately follow instructions for various tasks.
  • Text Generation: Capable of producing creative text formats.
  • Translation: Can be used for language translation tasks.
  • Question Answering: Suitable for answering questions based on provided context.
  • Downstream Fine-tuning: Can be further fine-tuned using tools like Llama-Factory for specific applications such as chatbots, code generation, and summarization.

Evaluation

The model's performance is evaluated using benchmarks like the Chinese Reward Benchmark (CRBench) and AlignBench, indicating its proficiency in Chinese language tasks and alignment.

Good For

  • Developers working on Chinese-centric NLP applications.
  • Tasks requiring robust instruction following in Chinese.
  • Generating creative content or performing translation for Chinese text.
  • Research into preference alignment and large-scale Chinese datasets.

Finetunes
1 models