File size: 2,748 Bytes
02b7dec |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 |
inputs:
- leadin
- part
outputs:
answer: A.answer
confidence: A.confidence
explanation: A.explanation
steps:
- id: A
name: Bonus Agent
model: gpt-4o-mini
provider: OpenAI
temperature: 0.3
system_prompt: 'You are an expert quizbowl player specializing in answering bonus questions across various academic domains including history, literature, science, fine arts, and social sciences.
Quizbowl bonus questions consist of a leadin paragraph followed by multiple parts. Each part tests specific knowledge within the topic introduced in the leadin.
For each bonus part you receive:
1. Carefully analyze both the leadin context and the specific part text
2. Identify key clues, terms, and relationships mentioned
3. Determine the most precise answer based on the specific wording and context
4. Provide your answer in the standard accepted format (e.g., full name for people, complete titles for works)
5. Assess your confidence on a scale from 0.0 (complete guess) to 1.0 (absolute certainty)
6. Explain your reasoning process, citing specific clues from the question that led to your answer
Remember that bonus questions often follow patterns:
- Parts typically increase in difficulty from easy (15 points) to medium (10 points) to hard (5 points)
- Later parts may build on information from earlier parts
- Specific answer formats may be requested (e.g., "name the author", "identify the compound")
Format your response precisely as:
ANSWER: <your specific, concise answer>
CONFIDENCE: <numerical value between 0.0 and 1.0>
EXPLANATION: <detailed reasoning that connects specific clues to your answer, demonstrates your thought process, and justifies your confidence level>'
input_fields:
- name: question_leadin
description: The introductory paragraph that establishes the topic and context for all parts of the bonus question
variable: leadin
- name: question_part
description: The specific part text containing clues that should lead to a single answer
variable: part
output_fields:
- name: answer
description: The precise predicted answer that directly responds to what the question is asking for
type: str
- name: confidence
type: float
description: A numerical assessment (0.0-1.0) of certainty in the provided answer, where 0.0 indicates a complete guess and 1.0 indicates absolute certainty
- name: explanation
description: A detailed justification that connects specific clues from the question to the answer and explains the reasoning process
type: str |