Publish Visual Scaffolding release files

DATA_LICENSE.md

@@ -0,0 +1,12 @@
+# Benchmark data license
+
+The benchmark image compositions and annotations are provided under the Creative Commons Attribution-ShareAlike 4.0 International license. The license text is included in [licenses/OpenMoji-CC-BY-SA-4.0.txt](licenses/OpenMoji-CC-BY-SA-4.0.txt).
+
+Upstream artwork retains its own attribution and license:
+
+- **OpenMoji**, by the OpenMoji contributors at HfG Schwäbisch Gmünd: https://openmoji.org/ and https://github.com/hfg-gmuend/openmoji. The artwork is licensed under CC BY-SA 4.0 and is used in Burden Counting.
+- **Twemoji**, by Twitter, Inc. and other contributors: https://github.com/twitter/twemoji. The graphics are licensed under CC BY 4.0, reproduced in [licenses/Twemoji-CC-BY-4.0.txt](licenses/Twemoji-CC-BY-4.0.txt), and are used for icon endpoints in Path Tracing.
+
+The benchmark resizes and composites graphics into generated scenes. Counting scenes can also rotate and occlude instances. Path scenes place icons at generated curve endpoints. These modifications form the benchmark compositions and do not imply endorsement by the artwork creators.
+
+Code and model weights are covered separately as described in the repository README.

NOTICE.md

@@ -0,0 +1,7 @@
+# Attribution
+
+Visual Scaffolding code: Copyright 2026 Visual Scaffolding contributors. Distributed under Apache License 2.0.
+
+The pretrained models are derived from Qwen3-VL-4B-Thinking by the Qwen team. Backbone source: https://huggingface.co/Qwen/Qwen3-VL-4B-Thinking. Backbone license: Apache License 2.0. The release adds learned visual retrieval and evidence adaptation modules and includes task-adapted model weights.
+
+Benchmark artwork includes OpenMoji and Twemoji. Their authors, source URLs, licenses, and transformations are identified in DATA_LICENSE.md. Their graphics licenses remain applicable to the incorporated artwork.

README.md

@@ -0,0 +1,85 @@
+---
+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.
+
+| 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.

SCHEMA.md

@@ -0,0 +1,34 @@
+# NMR-Bench test data
+
+`manifest.json` identifies each task's annotation file, sample count, and annotation-file SHA-256. Each `test.jsonl` record contains:
+
+| Field | Meaning |
+|---|---|
+| `sample_id` | Stable benchmark identifier |
+| `image` | Image path relative to the task directory |
+| `image_sha256` | SHA-256 of the supplied PNG |
+| `question` | Exact model input question |
+| `answer` | Canonical target answer, used only by the scorer |
+| `difficulty` | `easy`, `medium`, or `hard` |
+| `task_form` | Counting question type, when applicable |
+| `endpoint_type` | Path endpoint-marker category, when applicable |
+
+The corresponding `metadata.jsonl` in each task directory exposes the same records to the Hugging Face image viewer. It uses `file_name` instead of `image` to associate each row with a PNG under `images/`; `test.jsonl` remains the evaluator's canonical annotation file.
+
+Burden Counting has 420 images: seven forms × three difficulty levels × 20 images. Path Tracing has 360 images: three difficulty levels × 120 images, balanced across letters, numbers, icons, and shapes.
+
+| Counting form | Question | Boxed answer |
+|---|---|---|
+| T1a | Category count | Integer |
+| T2a | Category and size | Integer |
+| T2b | Category and tilt | Integer |
+| T2c | Category and occlusion | Integer |
+| T2d | Category, size, and tilt | Integer |
+| T3a | Compare size-group counts | Group label, count difference |
+| T3b | Most frequent listed category | Category label, count |
+
+Path answers list unordered endpoint pairs, for example `\boxed{(A, G), (B, M), (L, N)}`. Labels are normalized for case, numeric formatting, and backslash escapes.
+
+The last boxed answer is used. Counting also supports the scorer's reasoning-channel fallback when a model provides reasoning separately. In the single tied T3a case, the stored canonical answer is `small, 0`, applied consistently to all models.
+
+The model adapter receives only the image and question. Answers are reserved for scoring. Do not use target answers to construct prompts or predictions.