# VisToolBench Dataset

A benchmark dataset for evaluating vision-language models on tool-use tasks.

## Dataset Statistics

- **Total samples**: 1204
- **Single-turn**: 603
- **Multi-turn**: 601

## Schema

| Column | Type | Description |
|--------|------|-------------|
| `id` | string | Unique task identifier |
| `turncase` | string | Either "single-turn" or "multi-turn" |
| `num_turns` | int | Number of conversation turns (1 for single-turn) |
| `prompt_category` | string | Task category (e.g., "medical", "scientific", "general") |
| `eval_focus` | string | What aspect is being evaluated (e.g., "visual_reasoning", "tool_use") |
| `prompt` | string | The user prompt/question. For multi-turn, turns are prefixed with `[Turn N]` |
| `golden_answer` | string | The reference/ground-truth answer |
| `image` | Image | Primary image for the task (displayed in HF viewer) |
| `images` | List[Image] | All images associated with the task |
| `num_images` | int | Total number of images |
| `tool_trajectory` | string | JSON string of tool calls made (if applicable) |
| `rubrics` | string | JSON string of evaluation rubrics with weights and metadata |

## Rubrics Format

Each rubric entry contains:
- `description`: What the rubric evaluates
- `weight`: Importance weight (1-5)
- `objective/subjective`: Whether evaluation is objective or subjective
- `explicit/implicit`: Whether the answer is explicit or implicit in the image
- `category`: List of categories (e.g., "instruction following", "truthfulness")
- `critical`: Whether this is a critical rubric ("yes"/"no")
- `final_answer`: Whether this relates to the final answer ("yes"/"no")

## Usage

```python
from datasets import load_dataset

# Load the dataset
ds = load_dataset("path/to/dataset")

# Access a sample
sample = ds['train'][0]
print(sample['prompt'])
print(sample['image'])  # PIL Image

# Parse rubrics
import json
rubrics = json.loads(sample['rubrics'])
for rubric_id, rubric in rubrics.items():
    print(f"{rubric['description']} (weight: {rubric['weight']})")
```

## Splits

- `train`: Full dataset (1204 samples)

## Citation

```
@article{guo2025beyond,
  title={Beyond seeing: Evaluating multimodal llms on tool-enabled image perception, transformation, and reasoning},
  author={Guo, Xingang and Tyagi, Utkarsh and Gosai, Advait and Vergara, Paula and Park, Jayeon and Montoya, Ernesto Gabriel Hern{\'a}ndez and Zhang, Chen Bo Calvin and Hu, Bin and He, Yunzhong and Liu, Bing and others},
  journal={arXiv preprint arXiv:2510.12712},
  year={2025}
}
```