README.md

---
license: cc-by-sa-4.0
language:
- en
pretty_name: NMR-Bench
task_categories:
- image-text-to-text
tags:
- multimodal-reasoning
- visual-reasoning
- benchmark
size_categories:
- n<1K
configs:
- config_name: burden_counting
  data_files:
  - split: test
    path: burden_counting/images/*.png
  drop_labels: true
- config_name: path_tracing
  data_files:
  - split: test
    path: path_tracing/images/*.png
  drop_labels: true
---

# NMR-Bench

**A visual reasoning benchmark for [Visual Scaffolding](https://github.com/GeoX-Lab/VisualScaffolding).** It tests whether a model can keep track of object identity and spatial continuity while reasoning over an image.

📄 **Paper:** [Visual Scaffolding: Toward Native Multimodal Reasoning](https://arxiv.org/abs/2610.25247)

| Task | Challenge | Test images |
|---|---|---:|
| **Burden Counting** | Count and compare objects in crowded scenes | 420 |
| **Path Tracing** | Follow curves through crossings to match endpoints | 360 |

Task-specific checkpoints: [Burden Counting](https://huggingface.co/ShaoRun/vscaffold-4b-burden-counting) · [Path Tracing](https://huggingface.co/ShaoRun/vscaffold-4b-path-tracing).

## 🖼️ A look at the benchmark

Two **hard** examples from the test set are shown below. The queries and ground truths match the released annotations.

| Burden Counting | Path Tracing |
|:---:|:---:|
| <img src="burden_counting/images/counting_test_T2c_1000040.png" width="340" alt="Hard Burden Counting sample with partly occluded butterflies"> | <img src="path_tracing/images/path_test_cell060_seed00060000.png" width="340" alt="Hard Path Tracing sample with icon endpoints"> |

**Burden Counting** · `counting_test_T2c_1000040`

> **Query:** How many butterflies in this image are partially covered by another object? Give your final answer as a single integer inside `\boxed{}`, for example `\boxed{7}`.
>
> **GT:** There are `\boxed{21}` covered butterflies.

**Path Tracing** · `path_test_cell060_seed00060000`

> **Query:** The image contains multiple curves, each connecting two endpoints in the image. The endpoints in this image are labeled: {blue_circle, cat, crown, diamond_shape, dog, frog, hammer, heart, light_bulb, mushroom, star, sun}. Each label appears exactly once. Your task is to identify all connection pairs.
>
> Trace each curve carefully. Pay attention to crossings where lines overlap — at each crossing, the correct continuation should maintain smooth direction and curvature consistency. After completing all traces, verify your results to ensure they are consistent and correct.
>
> Output the connection pairs in the format: `\boxed{(label_a, label_b), (label_c, label_d), ...}`
>
> **GT:** Based on the traces above, the connection pairs are: `\boxed{(blue_circle, star), (cat, diamond_shape), (crown, heart), (dog, light_bulb), (frog, mushroom), (hammer, sun)}`.

## 🚀 Get started

With Python 3.12, download the test sets and run the scorer from the [GitHub repository](https://github.com/GeoX-Lab/VisualScaffolding):

```bash
git clone https://github.com/GeoX-Lab/VisualScaffolding.git
cd VisualScaffolding
pip install -e .
pip install huggingface_hub
hf download ShaoRun/nmr-bench --repo-type dataset --local-dir benchmark
python -m visual_scaffolding score \
  --task burden_counting \
  --benchmark benchmark \
  --predictions predictions/counting.jsonl \
  --output outputs/counting-summary.json
```

The scorer expects a JSONL prediction for every test sample, with `sample_id` and `response` fields. Use `--task path_tracing` for the other task. See the [GitHub README](https://github.com/GeoX-Lab/VisualScaffolding#readme) for model inference and the full evaluation workflow.

## 📂 Data and license

Each task has an `images/` directory, `metadata.jsonl` for the Hub's image viewer, and `test.jsonl` for the companion evaluator. Records contain the image path, question, answer, difficulty, and sample checksum. **Use the answer only for scoring; pass the image and question to the model.** Field definitions are in [SCHEMA.md](SCHEMA.md), and file checksums are in `manifest.json`.

Compositions and annotations are licensed under **CC BY-SA 4.0**. The images include OpenMoji and Twemoji artwork under their respective licenses. See [DATA_LICENSE.md](DATA_LICENSE.md) and [NOTICE.md](NOTICE.md) for attribution.