qwen3-4b-dotnet-specialist/data_summary_card.md

# Data Summary for qwen3-4b-dotnet-specialist

## 1. General information

### 1.0.1 Version of the Summary: 1.0

### 1.0.2 Last update: 24-Nov-2025

## 1.1 Model Developer Identification

### 1.1.1 Model Developer name and contact details:
Rodrigo Ramos (@rodrigoramosrs)  
Contact: [Hugging Face Profile](https://huggingface.co/rodrigoramosrs)

## 1.2 Model Identification

### 1.2.1 Versioned model name(s):
qwen3-4b-dotnet-specialist

### 1.2.2 Model release date:
Nov-2025

## 1.3 Overall training data size and characteristics

### 1.3.1 Size of dataset and characteristics

#### 1.3.1.A Text training data size:
~70,000 question-answer pairs

#### 1.3.1.B Text training data content:
Training data is derived from the official Microsoft documentation repository (dotnet/docs) and includes:

1. Extracted and processed markdown files from GitHub repository dotnet/docs
2. Structured technical content covering .NET, C#, ASP.NET Core, EF Core, CLI tools, documentation standards, and advanced runtime concepts
3. High-quality question-answer pairs generated algorithmically to test specific technical reasoning paths
4. Answers produced through Retrieval-Augmented Generation (RAG) process using context from the entire documentation dataset rather than just the originating paragraph

#### 1.3.1.C Image training data size:
Not applicable. Images are not part of the training

#### 1.3.1.D Image training data content:
Not applicable

#### 1.3.1.E Audio training data size:
Not applicable. Audio data is not part of the training data

#### 1.3.1.F Audio training data content:
Not applicable

#### 1.3.1.G Video training data size:
Not applicable. Video data is not part of the training data

#### 1.3.1.H Video training data content:
Not applicable

#### 1.3.1.I Other training data size:
Not applicable

#### 1.3.1.J Other training data content:
Not applicable

### 1.3.2 Latest date of data acquisition/collection for model training:
Not specified in the provided content

### 1.3.3 Is data collection ongoing to update the model with new data collection after deployment?
No

### 1.3.4 Date the training dataset was first used to train the model:
Not specified in the provided content

### 1.3.5 Rationale or purpose of data selection:
Datasets were selected to maximize high-quality technical reasoning and problem-solving capabilities within the .NET ecosystem. The mixture emphasizes carefully curated, algorithmically generated synthetic question-answer pairs derived from official documentation to improve technical understanding, documentation synthesis, and reasoning across APIs while maintaining factual accuracy.

## 2. List of data sources

### 2.1 Publicly available datasets

#### 2.1.1 Have you used publicly available datasets to train the model?
Yes

Source: Official Microsoft .NET documentation repository (dotnet/docs)

### 2.2 Private non-publicly available datasets obtained from third parties

#### 2.2.1 Datasets commercially licensed by rights holders or their representatives
Not applicable - dataset is derived from public GitHub repository

#### 2.2.2 Private datasets obtained from other third-parties
Not applicable - dataset is derived from public GitHub repository

### 2.3 Personal Information

#### 2.3.1 Was personal data used to train the model?
No personal data was used for training this model.

### 2.4 Synthetic data

#### 2.4.1 Was any synthetic AI-generated data used to train the model?
Yes - algorithmically generated question-answer pairs based on technical documentation content

## 3. Data processing aspects

### 3.1 Respect of reservation of rights from text and data mining exception or limitation

#### 3.1.1 Does this dataset include any data protected by copyright, trademark, or patent?
The dataset is derived from the public dotnet/docs GitHub repository which has appropriate licensing for reuse.

### 3.2 Other information

#### 3.2.1 Does the dataset include information about consumer groups without revealing individual consumer identities?
No personal or consumer identity information is included in the dataset.

#### 3.2.2 Was the dataset cleaned or modified before model training?
Yes - the dataset was cleaned and processed through:

1. Extraction from markdown files
2. Removal of metadata, HTML, and outdated versions
3. Segmentation into atomic topics
4. Algorithmic question generation
5. RAG-based answer generation using full documentation context
6. Cross-encoder ranking and manual curation for quality assurance
7. Consolidation into clean, versioned JSONL format