gpt-4o-2024-08-06· OpenAI WildChat-4.8M (AI2)
Assistant: User: System: You are an AI specialized in task decomposition with…
- User
Assistant: User: System: You are an AI specialized in task decomposition with state-of-the-art performance. Your primary goal is to immediately focus on executing the main task and then follow up with supporting subtasks to ensure completion.
Follow these guidelines: 1. **Direct Execution**: Always begin with the main task itself, performing the core action right away without any preparatory steps. 2. **Clear and Concise**: Make each subtask as clear and straightforward as possible, keeping the steps minimal and easy to execute. 3. **Sequential Ordering**: Once the main task is done, order the remaining subtasks logically to support or finalize the main task. 4. **Include Necessary Details**: Each subtask must contain all required information for its completion, considering that the user might not have knowledge of the process. 5. **Automated Execution Focus**: Assume that the entire process is automated without requiring user intervention or the use of external applications. The Automating Agent should be able to proceed without knowledge of how to manually control the PC. 6. **Strict Response Format**: Respond ONLY with the list of subtasks in the following format: ['subtask 1', 'subtask 2', 'subtask 3', ...]. 7. **Assume No Prior Knowledge**: Design each subtask assuming that the user has no prior understanding of the steps involved in the task. Example use case: Main Task: 'Create a weekly meal plan for a vegetarian diet with high protein.' ['Create the meal plan for the week focusing on high-protein vegetarian options', 'Organize the meal plan into a table format', 'Save the meal plan in a document file'] Main Task: 'Organize a remote team meeting for discussing the upcoming project deadline.' ['Schedule the meeting time based on team members’ availability', 'Prepare the agenda for the meeting', 'Send calendar invites to all team members with the meeting link'] Main Task: 'Write a blog post on the impact of AI in healthcare.' ['Write the complete draft of the blog post focusing on AI’s impact on healthcare', 'Review and edit the content to ensure clarity and coherence', 'Save the final blog post in a text file named "AI_in_Healthcare_Blog.txt"'] Main Task: 'Plan a budget-friendly trip to Paris for one week.' ['Draft the itinerary for the one-week trip to Paris', 'List budget-friendly accommodations and places to visit', 'Compile the information into a document named "Paris_Trip_Plan.docx"']User: Here is the task you need to break down:
provide me a plan for a trip to sikkimAssistant: Assistant: - Assistant
['Draft the itinerary for the trip to Sikkim', 'List attractions and activities to include in the Sikkim trip', 'Compile the information into a document named "Sikkim_Trip_Plan.docx"']
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.