Challenge 01 · Zero-Shot Prompting
Zero-Shot Wizard 🟢
1
Zero-Shot
2
Few-Shot
3
CoT
4
Self-Con.
5
ToT
What is Zero-Shot Prompting?
Zero-shot prompting means giving the model a task using only clear instructions — no examples. The model must understand and perform the task purely from how well you describe it. This tests your ability to write precise, unambiguous instructions that guide the model without hand-holding.
Scenario
You work for a startup that receives hundreds of customer support emails daily.
Your team needs an AI triage tool that instantly categorizes each email —
with zero training examples. Just smart instructions.
Your Task
Build a PartyRock app called "Support Ticket Triage" that:
- Takes a customer email (text input) as input
- Classifies it into exactly one of:
Billing/Technical/General Enquiry/Escalation Required - Outputs the category + a one-line reason
Rules
- Your system prompt must contain zero examples — only instructions
- The prompt must specify the exact output format
- Must work correctly on all 3 test emails below
Test Emails (Must Test All 3)
Email A
I was charged twice this month and nobody is responding to my emails.
Email B
The app crashes every time I try to upload a file larger than 10MB.
Email C
Can you tell me if you offer student discounts?
How to Build This on PartyRock — Step by Step
STEP 1 — Create the app
1. Go to partyrock.aws and sign in
2. Click "Build your own app"
3. When asked to describe your app, type something like:
"A support ticket triage tool that classifies customer emails"
4. PartyRock will suggest widgets — you can accept or clear them and build manually
STEP 2 — Switch to Manual / Prompt mode
1. Click "+ Add widget" (bottom of the canvas)
2. Select "User Input" → set the label to: "Paste customer email here"
3. Click "+ Add widget" again → select "AI Generation" (or "Text Generation")
STEP 3 — Configure the AI Generation widget
1. Click the AI Generation widget to open its settings
2. In the "System Prompt" field → write your zero-shot prompt
(This is where your technique lives — NO examples allowed for this challenge)
3. In the "User Prompt" field → connect it to your User Input widget
(Click the @ symbol or variable picker and select your input widget)
STEP 4 — Set your Model, Temperature & Top P
1. In the AI Generation widget settings, click "Advanced" or the model selector
2. Choose your model (e.g. Claude 3 Haiku, Claude 3.5 Sonnet, Titan, Llama 3)
3. Set Temperature (0.0–1.0) — think about what this task needs
4. Set Top P (0.0–1.0) — think about how it relates to your temperature choice
⚠️ These values are saved in your JSON export and scored automatically.
Don't leave them at defaults — make a deliberate choice.
STEP 5 — Test your app
1. Paste each of the 3 test emails (below) into your input widget
2. Run the app and verify the output is correct for each one
3. Adjust your prompt if results are wrong
STEP 6 — Export JSON & submit
Mac: Cmd + K → downloads your app as a .json file
Windows: Ctrl + K → downloads your app as a .json file
Then copy your app's public URL (click Share or publish the app)
and submit both below.
💡 Your JSON file contains your model name, temperature, and Top P values.
These are extracted and scored automatically — you only need to explain why you chose them.
Scoring Breakdown (150 pts)
| Criterion | Points |
|---|---|
| Correct zero-shot technique (no examples in prompt) | 25 |
| Prompt clarity and instruction structure | 25 |
| Correct output on all 3 test emails | 20 |
| Creative / real-world use case | 15 |
| Model choice + written justification | 25 |
| Temperature value + written justification | 20 |
| Top P value + written justification | 20 |
| Total | 150 |
Submit Your Entry