More Examples

April 1, 2026 · View on GitHub

This document lists more examples beyond those in the main README. To run all of them in one go, use examples/examples.jsonl with the --jsonl_path option (see the README section Test Multiple Questions in a Single Run).

Example 8

This example is from MindCube:

python example.py \
  --image_paths examples/Q8_1.jpg examples/Q8_2.jpg examples/Q8_3.jpg examples/Q8_4.jpg \
  --question "Based on these four images (image 1, 2, 3, and 4) showing the pink bottle from different viewpoints (front, left, back, and right), with each camera aligned with room walls and partially capturing the surroundings: From the viewpoint presented in image 4, what is to the left of the pink bottle?\nOptions: A. Pink plush toy and headboard B. Window and blue curtain C. Closet and door D. White wall\nAnswer with the option's letter from the given choices directly." \
  --model_path sensenova/SenseNova-SI-1.3-InternVL3-8B
Details of Example 8

Q: Based on these four images (image 1, 2, 3, and 4) showing the pink bottle from different viewpoints (front, left, back, and right), with each camera aligned with room walls and partially capturing the surroundings: From the viewpoint presented in image 4, what is to the left of the pink bottle?\nOptions: A. Pink plush toy and headboard B. Window and blue curtain C. Closet and door D. White wall\nAnswer with the option's letter from the given choices directly.

Image 1 Image 2 Image 3 Image 4

GT: C

Example 9

This example is from SITE-Bench:

python example.py \
  --image_paths examples/Q9.jpg \
  --question "Question: Consider the real-world 3D locations and orientations of the objects. Which side of the bus in the center is facing the bus stop?\nOptions: \nA. front\nB. left\nC. back\nD. right\nGive me the answer letter directly. The best answer is:" \
  --model_path sensenova/SenseNova-SI-1.3-InternVL3-8B
Details of Example 9

Q: Question: Consider the real-world 3D locations and orientations of the objects. Which side of the bus in the center is facing the bus stop?\nOptions: \nA. front\nB. left\nC. back\nD. right\nGive me the answer letter directly. The best answer is:

Image

GT: D

Example 10

This example is from SITE-Bench:

python example.py \
  --image_paths examples/Q10.jpg \
  --question "Question: Consider the real-world 3D orientations of the objects. Are the arrow on street sign and the taxi facing same or similar directions, or very different directions?\nOptions: \nA. same or similar directions\nB. very different directions\nGive me the answer letter directly. The best answer is:" \
  --model_path sensenova/SenseNova-SI-1.3-InternVL3-8B
Details of Example 10

Q: Question: Consider the real-world 3D orientations of the objects. Are the arrow on street sign and the taxi facing same or similar directions, or very different directions? Options: A. same or similar directions, B. very different directions. Give me the answer letter directly. The best answer is:

Image

GT: A

Example 11

This example is from SITE-Bench:

python example.py \
  --image_paths examples/Q11.jpg \
  --question "Question: What shape are all the men standing in?\nOptions: A. circle B. rectangle C. triangle D. square\nGive me the answer letter directly. The best answer is:" \
  --model_path sensenova/SenseNova-SI-1.3-InternVL3-8B
Details of Example 11

Q: Question: What shape are all the men standing in?\nOptions: A. circle B. rectangle C. triangle D. square\nGive me the answer letter directly. The best answer is:

Image

GT: A

Example 12

This example is from ViewSpatial-Bench:

python example.py \
  --image_paths examples/Q12.jpg \
  --question "From the perspective of this man who doesn't wear glasses, where is the man wearing glasses located beside him?\nOptions: A. left B. back-right C. front D. right\nAnswer with the option's letter from the given choices directly." \
  --model_path sensenova/SenseNova-SI-1.3-InternVL3-8B
Details of Example 12

Q: From the perspective of this man who doesn't wear glasses, where is the man wearing glasses located beside him? Options: A. left, B. back-right, C. front, D. right. Answer with the option's letter from the given choices directly.

Image

GT: A

Example 13

This example is from MMSI-Bench and test the model's capability in open-ended short-answer questions:

python example.py \
  --image_paths examples/Q13_1.png examples/Q13_2.png \
  --question "The iMac is in the northern part of the room. In which direction is the area where students do their homework?" \
  --model_path sensenova/SenseNova-SI-1.3-InternVL3-8B
Details of Example 13

Q: The iMac is in the northern part of the room. In which direction is the area where students do their homework?

First image Second image

GT: Northwest corner

Example 14

This example is from MMSI-Bench and test the model's capability in open-ended short-answer questions:

python example.py \
  --image_paths examples/Q14_1.png examples/Q14_2.png \
  --question "How many building models are captured in total in these two pictures?" \
  --model_path sensenova/SenseNova-SI-1.3-InternVL3-8B
Details of Example 14

Q: How many building models are captured in total in these two pictures?

First image Second image

GT: 4

Example 15

This example demonstrates the model's capability in solid geometry(Three views):

python example.py \
  --image_paths examples/Q15.png \
  --question "请将你的思考过程放在<think></think>标签内，并将你的最终答案放在<answer></answer>标签内。" \
  --model_path sensenova/SenseNova-SI-1.5-InternVL3-8B
Details of Example 15

Q: Enclose your thinking process in <think> </think> tags and your final answer in <answer> </answer>

First image

GT: B

Example 16

This example demonstrates the model's capability in solid geometry(Three views):

python example.py \
  --image_paths examples/Q16.png \
  --question "请将你的思考过程放在<think></think>标签内，并将你的最终答案放在<answer></answer>标签内。" \
  --model_path sensenova/SenseNova-SI-1.5-InternVL3-8B
Details of Example 16

Q: Enclose your thinking process in <think> </think> tags and your final answer in <answer> </answer>

First image

GT: C

Example 17

This example demonstrates the model's capability in solid geometry(3D graphic reasoning):

python example.py \
  --image_paths examples/Q17.png \
  --question "请将你的思考过程放在<think></think>标签内，并将你的最终答案放在<answer></answer>标签内。" \
  --model_path sensenova/SenseNova-SI-1.5-InternVL3-8B
Details of Example 17

Q: Enclose your thinking process in <think> </think> tags and your final answer in <answer> </answer>

First image

GT: C

Example 18

This example demonstrates the model's capability in solid geometry(Three views):

python example.py \
  --image_paths examples/Q18.png \
  --question "请将你的思考过程放在<think></think>标签内，并将你的最终答案放在<answer></answer>标签内。" \
  --model_path sensenova/SenseNova-SI-1.5-InternVL3-8B
Details of Example 18

Q: Enclose your thinking process in <think> </think> tags and your final answer in <answer> </answer>

First image

GT: A