Prompts de chain-of-thought y self-consistency

Ingeniería rápida con la API de OpenAI

Fouad Trad

Machine Learning Engineer

Chain-of-thought prompting

Hace que los LLM den pasos de razonamiento (pensamientos) antes de responder
Útil para tareas de razonamiento complejo
Ayuda a reducir errores del modelo

Imagen que muestra cómo un prompt chain-of-thought pide al modelo resolver un problema paso a paso, y la salida contiene el razonamiento de cada paso.

Chain-of-thought prompting

Prompt estándar para resolver una tarea de razonamiento

prompt = """Q: You start with 15 books in your collection. At the bookstore, you 
purchase 8 new books. Then, you lend 3 to your friend and 2 to your cousin. Later, 
you visit another bookstore and buy 5 more books. How many books do you have now? 
A: The answer is"""

print(get_response(prompt))

25 books

Chain-of-thought prompting

Chain-of-thought para resolver una tarea de razonamiento

prompt = """Q: You start with 15 books in your collection. At the bookstore, you 
purchase 8 new books. Then, you lend 3 to your friend and 2 to your cousin. Later, 
you visit another bookstore and buy 5 more books. How many books do you have now? 
A: Let's think step by step""" 
print(get_response(prompt))

Step 1: Start with the number of books in your collection: 15 books
Step 2: Purchase 8 new books at the bookstore: 15 + 8 = 23 books
Step 3: Lend 3 books to your friend: 23 - 3 = 20 books
Step 4: Lend 2 books to your cousin: 20 - 2 = 18 books
Step 5: Visit another bookstore and buy 5 more books: 18 + 5 = 23 books
Therefore, you have 23 books now.

Chain-of-thought con few‑shots

example = """
Q: The odd numbers in this group add up to an even number:  9, 10, 13, 4, 2.
A: Adding all the odd numbers (9, 13) gives 22. The answer is True.
"""

question = """
Q: The odd numbers in this group add up to an even number: 15, 13, 82, 7. 
A:
"""

prompt = example + question
print(get_response(prompt))

Adding all the odd numbers (15, 13, 7) gives 35. The answer is False.

Chain-of-thought vs. prompts multi‑paso

Prompts multi‑paso

Incluyen pasos dentro del prompt

Diagrama que muestra que un prompt multi‑paso da al modelo una serie de pasos secuenciales a ejecutar.

Chain-of-thought vs. prompts multi‑paso

Prompts multi‑paso

Incluyen pasos dentro del prompt

Diagrama que muestra que un prompt multi‑paso da al modelo una serie de pasos secuenciales a ejecutar.

Prompts chain-of-thought

Piden al modelo generar pasos intermedios

Imagen que muestra que en prompts chain-of-thought, los pasos los genera el modelo al producir la salida.

Limitación del chain-of-thought

Un pensamiento fallido --> resultado fallido
Se introdujeron los prompts self-consistency

Diagrama de un prompt chain-of-thought destacando que un pensamiento con razonamiento defectuoso lleva a un mal resultado.

Self-consistency prompting

Genera múltiples chain-of-thought repitiendo el prompt varias veces
Voto mayoritario para el resultado final

Imagen que muestra que un prompt self-consistency genera múltiples chain-of-thought, cada uno con salida, y el resultado final se obtiene por voto mayoritario.

Self-consistency prompting

Puede hacerse definiendo múltiples prompts o un prompt que genere múltiples respuestas.

self_consistency_instruction = "Imagine three completely independent experts who 
reason differently are answering this question. The final answer is obtained by 
majority vote. The question is: "


problem_to_solve = "If there are 10 cars in the parking lot and 3 more cars arrive. 
Half the original number of cars leave. Then, half of the current number of cars 
arrive. How many cars are there in the parking?"


prompt = self_consistency_instruction + problem_to_solve

print(get_response(prompt))

Prompt self-consistency

Expert 1: Let's go step by step [...] Therefore, the total number of cars in the 
parking lot is 8 + 4 = 12.

Expert 2: First, let's calculate [...] Therefore, the total number of cars in the 
parking lot is now 5 + 2 = 7 cars.

Expert 3: Initially, there are 10 cars [...] Thus, the final answer is 8 + 4 = 12 
cars in the parking lot.

Based on the majority vote, the final answer is that there are 12 cars in the 
parking lot.

¡Vamos a practicar!

Ingeniería rápida con la API de OpenAI