---
pretty_name: ActiveVision
license: cc-by-4.0
language:
- en
task_categories:
- visual-question-answering
- image-text-to-text
tags:
- benchmark
- visual-reasoning
- iterative-reasoning
- active-vision
- multimodal
- vqa
- evaluation
size_categories:
- n<1K
---
# ActiveVision — An Exam for Active Observers
**ActiveVision** is a benchmark for **iterative visual reasoning**: 85 photorealistic
items across **17 tasks** that cannot be solved from a single glance — the model has
to keep returning to the image to *scan*, *trace*, and *compare*. Every scene is
generated by a deterministic program and re-rendered photorealistically while
preserving the structure, so answers are exact by construction.
**Frontier models reach about 10%** with pure chain-of-thought (50.6% with an
autonomous coding agent), while **unaided humans average 96.1%**.
🌐 [Website](https://activevision.dev) · 📄 [Paper](https://arxiv.org/pdf/2607.16165) · 💻 [Evaluation code](https://github.com/saccharomycetes/ActiveVision)
## Examples
 |
 |
## The 17 tasks
| Family | Tasks |
|---|---|
| **Distributed Scanning** — find and count every signal | bounded_face_counting · connected_component_counting · region_counting · singleton_shape_counting · tangled_loop_counting |
| **Sequential Traversal** — follow a structure, step by step | arrow_chain_following · color_zone_sequencing · line_intersection_sequencing · maze_path_tracing · traversal_point_ordering |
| **Visual Attribute Transfer** — compare a property across regions | constellation_match_counting · contour_difference_spotting · field_difference_spotting · signal_difference_spotting · silhouette_match_counting · stroke_difference_spotting · stroke_match_counting |
5 items per task, 85 items total. Each item is **one still image and one question**;
answers are integers, letters, or ordered sequences — never small multiple choice, so
the guess floor is near zero.
## Dataset structure
```python
from datasets import load_dataset
ds = load_dataset("activevisionai/ActiveVision", split="train")
# columns: image, id, task, category, question, answer
```
| Field | Description |
|---|---|
| `image` | the photorealistic scene (PNG, up to 1536×1024) |
| `id` | item id, `-` (e.g. `MPT-2`) |
| `task` | one of the 17 task names above |
| `category` | task family: `distributed_scanning`, `sequential_traversal`, `visual_attribute_transfer` |
| `question` | the full question, instructing the model to answer in `...` tags |
| `answer` | ground truth (exact-match scoring, separator-insensitive) |
## Evaluation
Score the contents of the model's **last** `...` block by exact
match after normalizing case, whitespace, and separators (`"C, A, G"` ≡ `"CAG"`;
integer answers also match numerically, `"07"` ≡ `7`).
## Headline results (pure chain-of-thought, best per model)
| Model | Accuracy |
|---|---|
| Human (N=3, unaided) | **96.1%** |
| GPT-5.5 (xhigh) | 10.6% |
| Gemini 3.5 Flash (high) | 8.2% |
| Gemini 3.1 Pro (high) | 5.9% |
| Claude Opus 4.7 (max) | 4.7% |
| Claude Fable 5 (max) | 3.5% |
More reasoning does not close the gap: scaling effort moves models along the cost
axis, not toward the human band. Tool-using coding agents reach at most 50.6%
(Claude Code, Fable 5) at $2.74–$7.63 and 12–15 minutes per item — a human answers
unaided in 34 seconds.
## Citation
```bibtex
@article{zhang2026exam,
title={An exam for active observers},
author={Zhang, Jiarui and Tao, Muzi and Wang, Shangshang and Liu, Ollie and Ma, Xuezhe and Neiswanger, Willie},
journal={arXiv preprint arXiv:2607.16165},
year={2026}
}
```