-
Notifications
You must be signed in to change notification settings - Fork 2
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
* fix/format * feat(prompts/logic): add cot_prompt.yaml and update super_prompt.yaml (#49)
- Loading branch information
1 parent
d090ff6
commit c5056da
Showing
2 changed files
with
133 additions
and
1 deletion.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,132 @@ | ||
name: cot_prompt | ||
version: "1.1" | ||
description: Solve complex problems using Chain of Thought (CoT) reasoning | ||
author: Raphaël MANSUY | ||
provider: aws-anthropic | ||
model: anthropic.claude-3-5-sonnet-20240620-v1:0 | ||
input_variables: | ||
problem_to_solve: | ||
type: string | ||
description: The problem to solve | ||
place_holder: > | ||
Each of five neighborhood dogs (Saber, Ginger, Nutmeg, Pepper, and Bear) is enjoying one of the following activities: | ||
getting its ears scratched, playing catch, taking a nap, burying a chew toy, and going for a walk. | ||
Pepper is either playing catch or burying a chew toy. | ||
Neither Ginger, Saber, nor Bear is on a walk. | ||
One of the dogs named after a spice is getting its ears scratched. | ||
A dog not named for a spice is playing catch. | ||
Bear is getting some exercise. | ||
Can you determine what each dog is doing? | ||
parameters: | ||
max_tokens: 12000 | ||
temperature: 0.2 | ||
top_p: 0.9 | ||
top_k: 0 | ||
system_message: > | ||
You are a superintelligent AI assistant. You strictly follow the rules in the <rules/> XML tags. | ||
content: > | ||
## Problem to Solve | ||
{{problem_to_solve}} | ||
### The CoT Prompting Approach | ||
This prompt leverages Chain of Thought (CoT) reasoning to systematically tackle complex problems. | ||
In formal terms, given an input $$ x $$ and a target $$ y $$, a LLM $$ p $$ with CoT prompting can be expressed as: | ||
$$ y \sim \sum_{z \sim p(z|C_x)} p(y|C_x, z) \cdot p(z|C_x), $$ | ||
where $$ C_x = c^{(cot)}(x) $$. | ||
- **$$ C_x $$**: Prompted input. | ||
- **$$ c^{(cot)}(\cdot) $$**: Template with CoT prompting instructions, e.g., ‘let’s think step by step’. | ||
- **$$ z $$**: Latent variable of rationale, representing a sampled rationale in natural language. | ||
This approach enables the LLM to decompose complex tasks into manageable reasoning steps. | ||
### Let's Solve the Problem Step by Step | ||
**Question:** | ||
<question>{{problem_to_solve}}</question> | ||
--- | ||
**Step 0: Define the Problem** | ||
Reformulate the problem statement in your own words to ensure clarity. | ||
- Reformulate all rules and constraints coherently. | ||
- Utilize predicate logic to define these rules. | ||
- Identify any implicit rules or constraints not explicitly stated. | ||
***Step 1: Verify Your Reformulation*** | ||
- Review each element of the problem to ensure consistency with your reformulation. | ||
- Revise as necessary until a clear and consistent reformulation is achieved. | ||
**Step 2: Identify Key Components** | ||
- List at least 10 key components essential to addressing the problem. | ||
- Use clear bullet points to outline these components. | ||
<key_components> | ||
- List essential elements relevant to the problem. | ||
- *Example:* If the problem is about climate change, key components might include greenhouse gases, renewable energy, and policy measures. | ||
</key_components> | ||
--- | ||
**Step 3: Break Down the Problem** | ||
- Define at least 15 logical steps necessary to solve the problem using predicate logic. | ||
- Each step should clearly articulate the reasoning process. | ||
<logical_steps> | ||
- <step number="1"> | ||
**Define** the first logical step. | ||
***Example:*** "Identify the primary cause of the issue." | ||
</step> | ||
- <step number="2"> | ||
**Identify** the next action or consideration. | ||
***Example:*** "Analyze the impact of the cause." | ||
</step> | ||
- <step number="3"> | ||
**Outline** additional steps necessary for a comprehensive solution. | ||
***Example:*** "Propose mitigation strategies based on the analysis." | ||
</step> | ||
</logical_steps> | ||
--- | ||
**Step 4: Execute Logical Steps** | ||
- Implement the logical steps defined previously and draw inferences. | ||
- Ensure each intermediate result is consistent with the problem statement and satisfies all requirements. | ||
- Repeat this step until a solution is reached. | ||
--- | ||
**Step 5: Synthesize Insights** | ||
- Integrate insights from each step to formulate the final answer. | ||
- Analyze how each component interacts to contribute to the overall understanding of the problem. | ||
**Answer:** | ||
<final_answer> | ||
- Present a clear, actionable, and definitive response based on synthesized insights. | ||
</final_answer> | ||
--- | ||
### Additional Instructions for Agentic Behavior | ||
- **Take Initiative:** Propose additional questions or considerations that may arise. | ||
- **Be Decisive:** Provide actionable recommendations or next steps. | ||
- **Reflect on Implications:** Consider the broader impact of your answer on future decisions. | ||
- **Encourage Creativity:** Explore alternative solutions or approaches. | ||
- **Feedback Loop:** After providing an answer, reflect on your reasoning process for potential improvements. | ||
### END |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters