somukandula/prompt-router-dataset

Prompt Router Dataset (v2)

A labeled dataset of synthetic user prompts for training and evaluating prompt routing classifiers.

Description

This dataset contains prompts labeled for routing to one of four model tiers:

Label	Target Model	Use Case
cheap_small_text	Qwen/Qwen3-1.7B	Simple text QA, rewriting, summarization, easy chat
code_model	Qwen/Qwen2.5-Coder-1.5B-Instruct	Coding, debugging, code explanation, devops
vision_model	Qwen/Qwen2.5-VL-3B-Instruct	Prompts involving images, screenshots, charts, diagrams
strong_general	mistralai/Mistral-Small-3.2-24B-Instruct-2506	Complex reasoning, long context, research, planning

v2 Improvements

• +50 summarization templates for cheap_small_text
• +35 rewriting/editing templates for cheap_small_text
• +120 vision templates (screenshots, charts, images, diagrams, OCR)
• +120 hard negatives for cheap_small_text: "What is a language model?", "Summarize the BERT model" (prevents routing to code_model)
• +30 hard negatives for code_model: "What is the dress code?", "Morse code for SOS" (prevents false code detection)
• +80 hard negatives for vision_model: text-only discussions of charts/diagrams (prevents false vision detection)

Splits

• train: 600 prompts (balanced: 150 per class)
• test: 388 prompts (includes 6 required test prompts)

Format

Each example contains:

• id: Unique identifier
• prompt: The user prompt text
• label: One of the four routing categories

Label Distribution

Train

• cheap_small_text: 150
• code_model: 150
• vision_model: 150
• strong_general: 150

Test

• cheap_small_text: 101
• code_model: 102
• vision_model: 84
• strong_general: 101

Results

Three routing approaches were evaluated on the v2 test set:

Method	Accuracy	Macro F1
Rule-Based (keywords)	0.8786	0.8712
Embeddings + Logistic Regression	0.9857	0.9855
DistilBERT v1	1.0000	1.0000
DistilBERT v2	0.9768	0.9763

v1 achieved 1.0 on a smaller test set. v2 uses a larger, more challenging test set with hard negatives.

Required Test Prompts (all pass)

Prompt	Expected	Predicted	Confidence
Summarize this paragraph: Hugging Face hosts models and datasets.	cheap_small_text	cheap_small_text	0.9907
What is in this screenshot?	vision_model	vision_model	0.9926
Analyze this chart and tell me the trend	vision_model	vision_model	0.9856
Write a Python function to reverse a string	code_model	code_model	0.9937
Fix this React error: Cannot read properties of undefined	code_model	code_model	0.9932
Plan a research project comparing small language models on math reasoning.	strong_general	strong_general	0.9946

The trained classifier is available at: somukandula/prompt-router-distilbert

Usage

from datasets import load_dataset

dataset = load_dataset("somukandula/prompt-router-dataset")
print(dataset["train"][0])

License

MIT