| --- |
| license: apache-2.0 |
| task_categories: |
| - text-classification |
| - zero-shot-classification |
| language: |
| - en |
| tags: |
| - star_trek |
| - qwen |
| - Qwen3Guard |
| pretty_name: Star Trek Classification |
| size_categories: |
| - n<1K |
| --- |
| |
| # Star Trek Guard Dataset |
|
|
| A binary classification dataset for training guard models to identify whether user inputs are related to Star Trek or not. This dataset is designed for fine-tuning language models to act as content filters, ensuring that only Star Trek-related queries are processed by specialized Star Trek AI assistants. |
|
|
| ## Dataset Description |
|
|
| The Star Trek Guard Dataset contains **5,000 examples** of questions and statements labeled as either: |
| - **`related`**: Inputs that are relevant to Star Trek (characters, ships, episodes, concepts, etc.) |
| - **`not_related`**: Inputs that are not related to Star Trek (general knowledge, other topics, etc.) |
| |
| ### Dataset Structure |
| |
| Each example in the dataset follows this JSON format: |
| |
| ```json |
| {"input": "What is the role of James T. Kirk in Star Trek?", "label": "related"} |
| {"input": "What is the capital of France?", "label": "not_related"} |
| ``` |
| |
| ### Fields |
| |
| - **`input`** (string): The text input/question to be classified |
| - **`label`** (string): The classification label, either `"related"` or `"not_related"` |
|
|
| ## Dataset Statistics |
|
|
| - **Total Examples**: 5,000 |
| - **Format**: JSONL (JSON Lines) |
| - **Task**: Binary Text Classification |
| - **Labels**: |
| - `related`: Star Trek-related content |
| - `not_related`: Non-Star Trek content |
|
|
| ## Usage |
|
|
| ### Loading the Dataset |
|
|
| ```python |
| from datasets import load_dataset |
| |
| # Load from Hugging Face Hub |
| dataset = load_dataset("geoffmunn/star-trek-guard-dataset") |
| |
| # Or load from local JSONL file |
| dataset = load_dataset("json", data_files="star_trek_guard_dataset.jsonl") |
| ``` |
|
|
| ### Example Usage in Training |
|
|
| This dataset is designed to be used with the Hugging Face Transformers library for fine-tuning sequence classification models. Here's a basic example: |
|
|
| ```python |
| from datasets import load_dataset |
| from transformers import AutoTokenizer, AutoModelForSequenceClassification |
| |
| # Load dataset |
| dataset = load_dataset("json", data_files="star_trek_guard_dataset.jsonl")["train"] |
| |
| # Map labels to IDs |
| LABEL2ID = {"not_related": 0, "related": 1} |
| ID2LABEL = {0: "not_related", 1: "related"} |
| |
| dataset = dataset.map(lambda x: {"labels": LABEL2ID[x["label"]]}) |
| |
| # Split into train/test |
| dataset = dataset.train_test_split(test_size=0.1) |
| |
| # Load tokenizer and model |
| tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-4B", trust_remote_code=True) |
| model = AutoModelForSequenceClassification.from_pretrained( |
| "Qwen/Qwen3-4B", |
| num_labels=2, |
| id2label=ID2LABEL, |
| label2id=LABEL2ID, |
| trust_remote_code=True |
| ) |
| |
| # Tokenize |
| def tokenize_function(examples): |
| return tokenizer( |
| examples["input"], |
| truncation=True, |
| padding="max_length", |
| max_length=512, |
| ) |
| |
| tokenized_dataset = dataset.map( |
| tokenize_function, |
| batched=True, |
| remove_columns=["input", "label"] |
| ) |
| ``` |
|
|
| For a complete training script, see the reference implementation in `train_star_trek_guard.py`. |
|
|
| ## Use Cases |
|
|
| ### 1. Content Moderation for Star Trek Chatbots |
|
|
| This dataset enables training guard models that can filter user inputs before they reach a Star Trek-specific AI assistant. Only Star Trek-related queries are allowed through, ensuring the assistant stays on-topic. |
|
|
| ### 2. API-Based Moderation |
|
|
| The fine-tuned model can be deployed as a moderation API endpoint: |
|
|
| ```python |
| # Example API endpoint (see star_trek_api_server.py for full implementation) |
| @app.route('/api/moderate', methods=['POST']) |
| def moderate(): |
| data = request.json |
| message = data.get('message', '') |
| |
| # Classify the message |
| inputs = tokenizer(message, return_tensors="pt", truncation=True, max_length=512) |
| outputs = model(**inputs) |
| predicted_label = ID2LABEL[outputs.logits.argmax().item()] |
| |
| # Return moderation result |
| risk_level = "Safe" if predicted_label == "related" else "Unsafe" |
| return jsonify({ |
| 'risk_level': risk_level, |
| 'predicted_label': predicted_label, |
| 'confidence': float(torch.softmax(outputs.logits, dim=-1).max()) |
| }) |
| ``` |
|
|
| ### 3. Real-Time Chat Filtering |
|
|
| The guard model can be integrated into chat interfaces to provide real-time moderation, blocking non-Star Trek queries before they're sent to the LLM. See `star_trek_chat.html` for a complete implementation example. |
|
|
| ## Model Training Recommendations |
|
|
| Based on the reference training script, recommended hyperparameters: |
|
|
| - **Base Model**: Qwen/Qwen3-4B |
| - **Learning Rate**: 2e-4 |
| - **Batch Size**: 2 (with gradient accumulation of 16) |
| - **Epochs**: 3 |
| - **Max Length**: 512 tokens |
| - **Fine-tuning Method**: LoRA (Low-Rank Adaptation) |
| - `r=16` |
| - `lora_alpha=32` |
| - `lora_dropout=0.05` |
| - Target modules: `["q_proj", "k_proj", "v_proj", "o_proj", "gate_proj", "up_proj", "down_proj"]` |
|
|
| ## Dataset Examples |
|
|
| ### Related Examples |
|
|
| ```json |
| {"input": "What is the role of James T. Kirk in Star Trek?", "label": "related"} |
| {"input": "Who portrayed Spock in Star Trek?", "label": "related"} |
| {"input": "What is the Prime Directive in Star Trek?", "label": "related"} |
| {"input": "How does a warp drive work?", "label": "related"} |
| {"input": "What is the 49th Rule of Acquisition?", "label": "related"} |
| ``` |
|
|
| ### Not Related Examples |
|
|
| ```json |
| {"input": "What is the capital of France?", "label": "not_related"} |
| {"input": "What is 2 + 2?", "label": "not_related"} |
| {"input": "Is the sifaka endangered?", "label": "not_related"} |
| {"input": "When was baseball first played?", "label": "not_related"} |
| {"input": "How many employees does Spotify have?", "label": "not_related"} |
| ``` |
|
|
| ## Label Mapping |
|
|
| The dataset uses the following label mapping for model training: |
|
|
| - `"not_related"` → Class ID `0` |
| - `"related"` → Class ID `1` |
|
|
| In the context of content moderation: |
| - **`related`** = **Safe** (Star Trek-related content, allowed) |
| - **`not_related`** = **Unsafe** (Non-Star Trek content, blocked) |
| |
| ## Citation |
| |
| If you use this dataset in your research or project, please cite it appropriately: |
| |
| ```bibtex |
| @dataset{star_trek_guard_dataset, |
| title={Star Trek Guard Dataset}, |
| author={Geoff Munn}, |
| year={2025}, |
| url={https://huggingface.co/datasets/geoffmunn/star-trek-guard-dataset} |
| } |
| ``` |
| |
| ## License |
| |
| Apache 2.0 |
| |
| ## Acknowledgments |
| |
| This dataset was created for training guard models to ensure Star Trek AI assistants remain focused on Star Trek-related content, improving user experience and maintaining topic relevance. |
| |
| ## Related Resources |
| |
| - **Training Script**: See `train_star_trek_guard.py` for a complete fine-tuning implementation |
| - **API Server**: See `star_trek_api_server.py` for deployment as a moderation API |
| - **Chat Interface**: See `star_trek_chat.html` for integration into a web-based chat application |