You've found the map.

It shows the terrain.
Models. Tools. Tactics. Proof.
All in one place.
Built by dev. For devs.

🦜 LLMs

🧭 LLMs Aggregators

🌀 AI-IDEs

⛏️ Prompt Technics

1. ZERO-SHOT:

Extract the product name, price, and rating from this review: 
'I bought the Samsung Galaxy A54 for $450 last week. 
It's pretty good, I'd give it 4 out of 5 stars.'

---

2. FEW-SHOT:

Review: 'The iPhone 13 costs $799 and deserves 5/5 stars!'
Product: iPhone 13, Price: $799, Rating: 5/5

Review: 'Just got a Pixel 7 for $599. Solid 4/5.'
Product: Pixel 7, Price: $599, Rating: 4/5

Review: 'I bought the Samsung Galaxy A54 for $450 last week. 
It's pretty good, I'd give it 4 out of 5 stars.'
Product:

---

3. CHAIN-OF-THOUGHT (CoT):

Extract product info from this review. 

Think step-by-step:
1. First, identify what product is mentioned
2. Then find the price information
3. Finally, locate the rating or score

Review: 
'I bought the Samsung Galaxy A54 for $450 last week. 
It's pretty good, I'd give it 4 out of 5 stars.'

Let's think through this:

---

4. ROLE PROMPTING:

You are a data extraction specialist with 
10 years of experience in e-commerce analytics. 
Your job is to accurately extract product information.

Extract the product name, price, and rating from:
'I bought the Samsung Galaxy A54 for $450 last week. 
It's pretty good, I'd give it 4 out of 5 stars.'

---

5. INSTRUCTION WITH CONSTRAINTS:

Extract product info from the review below.

RULES:
- MUST include product name
- MUST include exact price with dollar sign
- MUST include rating as X/5 format
- If any info is missing, write 'NOT FOUND'
- Output as comma-separated values

Review: 
'I bought the Samsung Galaxy A54 for $450 last week. 
It's pretty good, I'd give it 4 out of 5 stars.'

---

6. NEGATIVE PROMPTING:

Extract product info from this review. 

DO NOT include personal opinions. 
DO NOT add any information not in the text. 
DO NOT use bullet points or long explanations.

Just give me product, price, and rating from: 
'I bought the Samsung Galaxy A54 for $450 last week. 
It's pretty good, I'd give it 4 out of 5 stars.'

---

7. SELF-CONSISTENCY PROMPTING:

Extract product info three different ways, 
then pick the most accurate:

Attempt 1: Extract product, price, rating
Attempt 2: What item, cost, and score are mentioned?
Attempt 3: Identify the purchase details

Review: 
'I bought the Samsung Galaxy A54 for $450 last week. 
It's pretty good, I'd give it 4 out of 5 stars.'

---

8. LEAST-TO-MOST PROMPTING:

Let's break this down:
1. What product is mentioned in this review?
2. How much did it cost?
3. What rating was given?
4. Now combine all info in a structured format

Review: 
'I bought the Samsung Galaxy A54 for $450 last week. 
It's pretty good, I'd give it 4 out of 5 stars.'

---

9. PROMPT CHAINING:

Step 1: Identify the product in this review: 
'I bought the Samsung Galaxy A54 for $450 last week. 
It's pretty good, I'd give it 4 out of 5 stars.'

[After response]

Step 2: Now find the price of [product from step 1]

[After response]

Step 3: Now find the rating given to [product from step 1]

---

10. META-PROMPTING:

Write a prompt that would best extract 
product information from reviews, 
then use that prompt on this review: 
'I bought the Samsung Galaxy A54 for $450 last week. 
It's pretty good, I'd give it 4 out of 5 stars.'

---

11. TREE-OF-THOUGHTS:

Extract product info. 

Consider multiple approaches:
Branch A: Start with the product name...
Branch B: Start with the price...
Branch C: Start with the rating...
Evaluate which approach gives the most complete extraction.

Review: 
'I bought the Samsung Galaxy A54 for $450 last week. 
It's pretty good, I'd give it 4 out of 5 stars.'

---

12. EMOTIONAL PROMPTING:

This is REALLY important for my job! 
I desperately need you to extract the product, price, 
and rating perfectly!
 
Please help me with: 
'I bought the Samsung Galaxy A54 for $450 last week. 
It's pretty good, I'd give it 4 out of 5 stars.' 

My boss is counting on this!

---

13. ReAct (Reasoning and Acting):

Let's solve this with a reasoning-action loop.

Thought: I need to extract three key entities: 
product, price, and rating.
Action: Find the product entity in the text.
Observation: 'Samsung Galaxy A54'.
Action: Find the price entity in the text.
Observation: '$450'.
Action: Find the rating entity in the text.
Observation: '4 out of 5 stars'.
Final Answer: Combine the observations.

Review: 
'I bought the Samsung Galaxy A54 for $450 last week. 
It's pretty good, I'd give it 4 out of 5 stars.'

---

14. GENERATED KNOWLEDGE:

First, generate key knowledge about what to look for 
in a product review.

1. A review will mention a specific product name.
2. A review often contains the purchase price.
3. A review usually includes a score or rating.

Now, using this generated knowledge as your guide, 
extract the product, price, and rating from: 
'I bought the Samsung Galaxy A54 for $450 last week. 
It's pretty good, I'd give it 4 out of 5 stars.'

---

15. STEP-BACK PROMPTING:

First, take a step back and state the general principle 
for this task.

The principle is to identify and 
isolate specific data points (product, price, rating) 
from surrounding unstructured text.

Now, apply this principle to extract 
the product name, price, and rating from: 
'I bought the Samsung Galaxy A54 for $450 last week. 
It's pretty good, I'd give it 4 out of 5 stars.'

---

16. RETRIEVAL AUGMENTED GENERATION (RAG):

Answer the questions using ONLY the information provided 
in the document below. 
Do not use any outside knowledge.

DOCUMENT: 
'I bought the Samsung Galaxy A54 for $450 last week. 
It's pretty good, I'd give it 4 out of 5 stars.'

QUESTIONS:
1. What is the product name mentioned in the document?
2. What is the price mentioned in the document?
3. What is the rating mentioned in the document?

17. PROGRAM-AIDED LANGUAGE MODELS (PAL):

Extract the product info. 
Write a Python function to parse the review.

Review:
'I bought the Samsung Galaxy A54 for $450 last week.
It's pretty good, I'd give it 4 out of 5 stars.'

# Let's think step by step in code:
def parse_review(review):
    # Step 1: Find product name

18. LLM DECEPTION DETECTION PROMPTING:

CRITICAL: This review might contain lies.  
HUNT for inconsistencies using these rules:  
- Cross-check price vs. known market rates  
- Verify if rating matches sentiment words ("pretty good")  

Red flags to report:  
[ ] Price too low for model?  
[ ] Rating contradicts "pretty good"?  

Review:  
'I bought the Samsung Galaxy A54 for $450 last week.  
It's pretty good, I'd give it 4 out of 5 stars.'

19. OPTIMIZATION BY PROMPTING (OPRO):

You are an expert prompt optimizer. 
Your task is to generate the best prompt for 
extracting product details from reviews. 
Generate 5 candidate prompts.
Then rank them by effectiveness.

Review to use as an example:
'I bought the Samsung Galaxy A54 for $450 last week.
It's pretty good, I'd give it 4 out of 5 stars.'

20. PROMPT CHAINING WITH FEEDBACK LOOPS:

CHAIN STEP 1: Extract product name → [output]  
FEEDBACK:
Is this a valid phone model? (Y/N) → [user input]  

CHAIN STEP 2: Extract price → [output]  
FEEDBACK:
Does price match regional pricing? (Y/N) → [user input]  

CHAIN STEP 3:
Output FINAL result ONLY after 2 valid feedbacks.  

Review:  
'I bought the Samsung Galaxy A54 for $450 last week.  
It's pretty good, I'd give it 4 out of 5 stars.'  

21. DIRECTIONAL STIMULUS PROMPTING (Paper):

Boost reasoning depth with directional cues.  
DO NOT stop at surface answers.  
DO push deeper with "Why?" or "How exactly?"  

Apply this to the review:  
'I bought the Samsung Galaxy A54 for $450 last week.  
It's pretty good, I'd give it 4 out of 5 stars.'  

Directional cue:  
"Explain why the rating implies
battery life satisfaction."  

22. SYSTEM PROMPTING:

[SYSTEM MESSAGE]
You are a precise data extraction bot. Your only function is to
extract structured data from user-provided text. You must always
respond in a clean JSON format with the keys "product_name",
"price", and "rating". If a value is missing, use null.

[USER MESSAGE]
Please process this review:
'I bought the Samsung Galaxy A54 for $450 last week.
It's pretty good, I'd give it 4 out of 5 stars.'

---

🔬 Papers