가이드 연계 퀴즈 · 중간 레벨
시각적 추론 퀴즈
이미지 인식 대 추론, 차트 척도 및 단위, 제어되고 사실적인 장면, 작은 텍스트, 언어 사전 및 증거 확인을 테스트합니다.
질문 1 의 8
A chart answer uses correct subtraction but reads the vertical axis incorrectly. Which stage failed?
모든 질문 미리보기8 질문
퀴즈를 풀면 정답과 설명이 나타납니다.
- A chart answer uses correct subtraction but reads the vertical axis incorrectly. Which stage failed?
- Arithmetic after the values were extracted
- The format of the final sentence
- The choice of image file extension
- Perception of the visual evidence before inference
- A chart axis starts at 90; bars mark 95 and 100. What is their numerical difference?
- Ten units
- Five units
- Ninety-five units
- A difference cannot be read from a nonzero axis
- A model names every object in an image. What ability remains unproven?
- Recognition of object labels
- Production of a sentence
- Reliable counting, spatial comparison and relationship inference
- Detection of color differences
- Why pair controlled diagnostic scenes with realistic images in evaluation?
- A narrow skill test may not transfer to cluttered real material
- Simple scenes always prove deployment quality
- Real scenes make ground truth impossible
- Controlled data replace every task check
- A model misreads small chart text but reads large headings. Which input factor should be inspected?
- Only the model’s English vocabulary size
- Resolution and cropping of the relevant labels
- Only the chart’s publication date
- Only the answer’s punctuation
- An image contradicts the most common answer to a question. Why is this a useful test?
- It removes the need for source images
- It guarantees all visual reasoning will succeed
- It makes the image caption a legal record
- It probes whether a language prior overrides actual visual evidence
- The model gives a detailed explanation of a chart comparison. What should a reviewer do?
- Check stated labels, values and relationships against the chart
- Assume detailed reasoning proves perception
- Verify only grammar
- Ignore the underlying image
- For consequential chart use, what record helps reconstruct the answer?
- Only the final natural-language paragraph
- Only a confidence adjective
- The source image and extracted values used in comparison
- Only a model name