Model Card for Qwen3KeywordsFangwusha06B
Qwen3KeywordsFangwusha06B is a fine-tuned language model based on Qwen3, optimized for Chinese keyword extraction, keyphrase generation, and structured text mining tasks in domain-specific scenarios.
Model Details
Model Description
This model is a 0.6B-parameter Chinese language model fine-tuned for keyword extraction and key information summarization tasks. It is built on the Qwen3 base model and further adapted to improve accuracy and recall in professional text keyword mining.
- Developed by: Yougen Yuan
- Funded by [optional]: Personal Research Project
- Shared by [optional]: Yougen Yuan
- Model type: Language Model, Keyword Extraction Model
- Language(s) (NLP): Chinese (Simplified)
- License: Apache-2.0
- Finetuned from model [optional]: Qwen3-0.6B
Model Sources [optional]
- Repository: https://huggingface.co/Yougen/Qwen3KeywordsFangwusha06B
- Paper [optional]: [More Information Needed]
- Demo [optional]: [More Information Needed]
Uses
Direct Use
This model can be directly used for:
- Chinese text keyword extraction
- Keyphrase generation from articles, documents, and paragraphs
- Lightweight text tagging and topic labeling
- Rapid key information extraction in low-resource deployment scenarios
Downstream Use [optional]
Can be further fine-tuned for:
- Domain-specific keyword extraction (news, academic, technical documents)
- Search engine keyword suggestion
- Content classification and tagging systems
- Integration into RAG systems for query and document keyphrase matching
Out-of-Scope Use
- Not designed for general dialogue generation or long-form text creation
- Not suitable for high-stakes decision-making systems without human review
- Not intended for harmful content generation, misinformation, or privacy-violating applications
- Not optimized for non-Chinese languages
Bias, Risks, and Limitations
- The model may inherit linguistic and domain biases present in the pre-training data of the base Qwen3 model.
- Performance may degrade on extremely short, noisy, or highly colloquial text.
- Keyword extraction results may vary across different domains and require post-processing for production use.
- The model does not have built-in fact-checking ability and may generate inappropriate keywords for sensitive content.
Recommendations
Users should manually validate key output results in production environments.
It is recommended to apply filtering rules for sensitive or low-quality text inputs.
Downstream developers should conduct domain-specific evaluation before deploying to real-world systems.
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model.
How to Get Started with the Model
Use the following code to load and infer with the model:
1from transformers import AutoTokenizer, AutoModelForCausalLM
2
3model_name = "Yougen/Qwen3KeywordsFangwusha06B"
4tokenizer = AutoTokenizer.from_pretrained(model_name)
5model = AutoModelForCausalLM.from_pretrained(model_name)
6
7inputs = tokenizer("提取文本关键词：", return_tensors="pt")
8outputs = model.generate(**inputs, max_new_tokens=64)
9print(tokenizer.decode(outputs[0], skip_special_tokens=True))
Training Details
Training Data
Training data consists of Chinese text-keyword paired datasets, including general articles, web text, and domain-specific annotated documents. Data preprocessing includes deduplication, noise filtering, and formatting into instruction-tuning style.
Training Procedure
Preprocessing [optional]
- Text cleaning and special symbol removal
- Instruction template formatting for keyword extraction
- Fixed-length token truncation and padding
Training Hyperparameters
- Training regime: bf16 mixed precision
- Learning rate: 2e-5
- Batch size: 16
- Epochs: 3
- Optimizer: AdamW
Speeds, Sizes, Times [optional]
- Model parameter size: 0.6B
- Training hardware: NVIDIA GPU
- Training time: Several hours
Evaluation
Testing Data, Factors & Metrics
Testing Data
Internal Chinese keyword extraction test set covering general and domain-specific texts.
Factors
Text length, domain type, noise level, keyword density.
Metrics
- Precision
- Recall
- F1-score
Results
[More Information Needed]
Summary
The model achieves stable performance in Chinese keyword extraction tasks with efficient inference speed suitable for edge and lightweight deployment.
Model Examination [optional]
[More Information Needed]
Environmental Impact
Carbon emissions can be estimated using the
Machine Learning Impact calculator presented in
Lacoste et al. (2019).
- Hardware Type: NVIDIA GPU
- Hours used: [More Information Needed]
- Cloud Provider: [More Information Needed]
- Compute Region: [More Information Needed]
- Carbon Emitted: [More Information Needed]
Technical Specifications [optional]
Model Architecture and Objective
Decoder-only transformer architecture based on Qwen3.
Optimized for instruction-following and sequence labeling-style keyword generation objectives.
Compute Infrastructure
Hardware
NVIDIA GPU with CUDA support
Software
- PyTorch
- Transformers
- Accelerate
- Datasets
Citation [optional]
BibTeX:
[More Information Needed]
APA:
[More Information Needed]
Glossary [optional]
- Keyword Extraction: The task of automatically identifying important terms from a text.
- Qwen3: Base large language model developed by Alibaba Cloud.
More Information [optional]
For questions and updates, please refer to the model repository on Hugging Face Hub.
Model Card Authors [optional]
Yougen Yuan
Model Card Contact
[More Information Needed]
