Node 3: <S4>
Node 4: <S5>
Output:
Figure 9: The prompt for causal relationships extraction.
1-Needle Question Generation (GPT-o1-preview-2024-09-12)
Caulsal 1-Needle Question
Given a pair of sentences with a causal relationship, create a question about the “Effect" sentence such that the answer to this question
is found exclusively in the “Cause" sentence. Ensure that the answer does not appear within the “Effect” sentence itself.
Example Input:
Cause: Blaming himself, Harry Hart, codenamed Galahad, delivers a medal for valor to Lee’s widow, Michelle and his young son, Eggsy,
saying that if they ever need help, they should call the phone number on the back of the medal.
Effect: Arrested for stealing a car, Eggsy calls the number on the medal.
Example Output:
Why did Eggsy have a phone number to call when he was arrested for stealing a car?
(continue with another four demonstrations)
Input:
Cause: <Cause Sentence>
Effect: <Effect Sentence>
Output:
.
Noncausal 1-Needle Question
Given a sentence, please use the content of it as context to create a question whose answer appears in the sentence.
Please output the question and answer in the following format, keeping the answer as concise as possible:
Question: <Question>
Answer: <Answer>
Input:
Sentence: <Cause Sentence/Effect Sentence>
Output:
Figure 10: The prompt for generating 1-needle questions.
Visual Grounding 2-Needle Question Generation (Step 1)
(Gemini-1.5-pro-002)
Step 1: Rephrase the cause and effect sentences
You have a pair of sentences: one indicating a cause and the other describing its effect.
Each sentence has a related image that complements the content expressed in the sentence.
Your goal is to rephrase both sentences so that the causal relationship between them is explicitly clear.
Please ensure that after rephrasing, there is a piece of shared information present in both sentences that establishes the causal relationship between them.
Remember that the rephrased sentences should not include any content that is not present in the original text or images.
Example 1:
Input:
Cause: Gordon, who was actually sent to save Rachel, is unable to make it there in time and Rachel dies.
<insert cause scene here>
Effect: The Joker is able to locate Dent in a hospital and manipulates him into seeking revenge for Rachel’s death.
<insert effect scene here>
Output:
Cause: Gordon, tasked with rescuing Rachel, arrives too late, and she is dead.
Effect: The Joker confronts Dent in a hospital and manipulates him into seeking revenge for Rachel’s death against Gordon.
Example 2:
Input:
Cause: The remaining robber reveals himself to be the joker, a crazed supervillain, and escapes the bank in a school bus.
<insert cause scene here>
Effect: Several mob leaders hold a conference, which is interrupted by the joker.
<insert effect scene here>
Output:
Cause: After the final robber reveals himself to be the joker—an unstable supervillain—and escapes the bank by blending into a line of school buses, the criminal underworld is thrown into disorder.
Effect: To restore order, several prominent mob leaders convene a clandestine meeting, only to have the joker himself crash their conference.
Input:
Cause: <Cause Sentence> <Cause Scene Images>
Effect: <Effect Sentence> <Effect Scene Images>
Figure 11: The prompt for generating visual grounding 2-needle questions.
Visual Grounding 2-Needle Question Generation (Step 2)
(GPT-o1-preview-2024-09-12)
Step 2: Generate questions based on bridge entities
You have a pair of sentences: one indicating a cause and the other describing its effect. Your goal is to create a sentence containing two questions. Please complete this goal step by step, as shown in the examples. (Please follow the output format shown in the example.)
Example Input:
Cause: Feeling responsible, Harry Hart, known as Galahad, gives a medal for valor to Lee’s widow, Michelle, and his young son, Eggsy, instructing them to call the number on the medal if they ever need help.
Effect: When Eggs