p_id stringclasses 10
values | prompt stringclasses 8
values | image imagewidth (px) 277 512 | gold_answer stringclasses 10
values | perturbed_answer stringclasses 10
values |
|---|---|---|---|---|
vg_ido_2 | Which option describe the object relationship in the image correctly? | Based on a detailed observation of the image, the primary object relationship is that the **cat is resting on and against the bag**.
Here is a breakdown of the specific visual details defining their relationship:
* **Positioning:** A dark grey tabby cat with white paws is lying on its side on a light-colored carpet.... | Based on a detailed observation of the image, here is a breakdown of the specific visual details:
* **Positioning:** A dark grey tabby cat with white paws is lying on its side on a light-colored carpet. It is positioned immediately adjacent to a dark blue and beige duffel bag.
* **Physical Contact:** The cat's body... | |
vg_ido_5 | What type of environment is depicted in the picture? | Based on the visual evidence in the image, the environment depicted is the interior of a **commercial gym, fitness center, or weightlifting facility**, specifically focusing on a free-weight training area.
Here is a detailed breakdown of the elements that define this environment:
**Foreground (Left and Bottom):**
* ... | Based on the visual evidence in the image, here is a detailed breakdown of the elements present:
**Foreground (Left and Bottom):**
* **Weight Plates:** The most prominent feature is the tall, organized stacks of thick, black weight plates dominating the left side of the frame. These appear to be "bumper plates," typ... | |
vg_ido_6 | Which mood does this image convey? | Based on the visual details in the image, the primary moods conveyed are **self-love, deep contentment, comfort, and peacefulness.**
Here is a detailed breakdown of the visual elements that contribute to this mood:
* **Body Language (The Self-Hug):** The most prominent feature of the image is the man wrapping his ... | Based on the visual details in the image, here is a detailed breakdown of the visual elements:
* **Body Language (The Self-Hug):** The most prominent feature of the image is the man wrapping his arms tightly around his own torso, with his hands gripping his opposite shoulders/upper arms. This gesture of hugging ones... | |
vg_ido_8 | Which group has more items? | Based on a careful examination of the image, the **left group** has more items.
Here is a detailed breakdown of what is visible in the image:
The image is divided into two separate rectangular panels, a left side and a right side, both containing illustrations of pink cartoon pigs of varying sizes.
* **Left Group:... | Based on a careful examination of the image, here is a detailed breakdown of what is visible:
The image is divided into two separate rectangular panels, a left side and a right side, both containing illustrations of pink cartoon pigs of varying sizes.
* **Left Group:** If you count the pigs in the left panel, there... | |
vg_ido_14 | Which one is the correct caption of this image? | Based on the detailed visual elements in the image, here is a comprehensive descriptive caption:
**A young person with short, dark hair stands in profile in a dimly lit room, washing something in a white basin.**
Here is a breakdown of the specific details that make up this caption:
* **The Subject:** In the cente... | Based on the detailed visual elements in the image, here is a comprehensive descriptive caption:
**A young person with short, dark hair stands in profile in a dimly lit room, washing something in a white basin.**
Here is a breakdown of the specific details that make up this caption:
* **The Subject:** In the cente... | |
vg_ido_17 | what is the shape of this object? | Based on the image provided, the overall shape of the primary object (the road sign) is a **triangle**.
To provide a more detailed and comprehensive breakdown of the shapes visible on the object:
* **Overall Outline:** It is specifically an upward-pointing **equilateral triangle**, meaning it has three sides of equ... | Based on the image provided, the primary object (the road sign) has a distinct geometric form.
To provide a more detailed and comprehensive breakdown of the shapes visible on the object:
* **Overall Outline:** It is an upward-pointing shape, meaning it has three sides of equal length and a horizontal base with one ... | |
vg_ido_19 | Which of the following captions best describes this image? | Based on the image, since no options were provided, here is a caption that best describes the scene, followed by a detailed breakdown of why it is accurate:
**Best Caption:** "A family of three wearing helmets while enjoying a bicycle ride on a paved trail on a sunny day."
**Detailed Breakdown of the Image:**
* **... | Based on the image, since no options were provided, here is a detailed breakdown of the scene:
* **The Subjects:** The image features three people who appear to be a family. On the left is a woman wearing a green plaid shirt and blue jeans. In the middle is a young boy wearing a bright red t-shirt and blue jeans. On... | |
vg_ido_20 | Which one is the correct caption of this image? | Based on the image provided, a correct and comprehensive caption would be:
**"A classic, two-tone turquoise and dark blue station wagon parked on grass, featuring whitewall tires, a lowered stance, and a matching surfboard secured to a wooden roof rack."**
Here is a detailed breakdown of the visual elements in the im... | Based on the image provided, here is a detailed breakdown of the visual elements that would support a correct and comprehensive caption:
* **The Vehicle:** The central focus is a vintage station wagon, likely from the mid-1950s. The twin silver streaks running down the hood are a hallmark design feature of Pontiacs ... | |
vg_ido_23 | There is a cyan metal thing behind the cyan metal ball; what shape is it? | Based on the image, the cyan metallic object located behind and to the left of the cyan metal ball is a **cylinder**. | Based on the image, there is a cyan metallic object located behind and to the left of the cyan metal ball. | |
vg_ido_24 | What type of environment is depicted in the picture? | Based on the visual evidence in the image, the environment depicted is a **surreal, magical fantasy landscape**, often associated with fairy tales or dreamscapes.
Here is a detailed breakdown of the environmental elements that contribute to this classification:
**1. The Foreground (The Portal/Frame):**
* **Intertw... | Based on the visual evidence in the image, here is a detailed breakdown of the environmental elements present:
**1. The Foreground (The Portal/Frame):**
* **Intertwined Flora:** The viewer is looking through a massive, near-perfect circular frame made of thick, gnarled tree trunks, roots, and vines. These wooden ele... |
Focus Dataset
Focus is a perturbation benchmark for evaluating VLM-as-judge systems.
Subsets
| Subset | Description | Splits |
|---|---|---|
i2t |
Image-to-Text perturbations | visual_grounding, semantic_interpretation, visual_reasoning, long_form_generation, score_invariant |
t2i |
Text-to-Image perturbations | visual_fidelity, scene_coherence, physical_plausibility, text_rendering, score_invariant |
Usage
from datasets import load_dataset
# Load a specific split
ds = load_dataset("sanjay73/Focus-test", "i2t", split="visual_grounding")
# Load all splits for a task
ds = load_dataset("sanjay73/Focus-test", "i2t")
Schema
I2T (Image-to-Text)
| Field | Type | Description |
|---|---|---|
p_id |
string | Perturbation ID |
prompt |
string | Question about the image |
image |
image | Input image |
gold_answer |
string | Correct answer |
perturbed_answer |
string | Perturbed (incorrect) answer |
T2I (Text-to-Image)
| Field | Type | Description |
|---|---|---|
p_id |
string | Perturbation ID |
prompt |
string | Text prompt |
gold_image |
image | Correctly generated image |
perturbed_image |
image | Perturbed (edited) image |
edit_instruction |
string | Edit applied to create perturbation |
- Downloads last month
- 54