> ## Documentation Index
> Fetch the complete documentation index at: https://docs.learningcommons.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Graphics Complexity

> Access informational passages with charts, diagrams, and illustrations from the Graphics Complexity evaluator's test set, each with a rating that literacy experts agreed with.

<EarlyAccess isBreaking={false} />

<Note>
  [Look at the
  data.](https://github.com/learning-commons-org/evaluators/tree/main/evals/graphics/ela/graphics-complexity) ↗
</Note>

## About the dataset

The Graphics Complexity dataset contains passages used to test the [Graphics Complexity evaluator](/evaluators/graphics-evaluators/introduction). It only includes passages where a majority of literacy experts agreed with the evaluator's rating. Each passage is an informational text with one or more graphics, such as photographs, illustrations, diagrams, charts, or maps. The evaluator didn't see any of these passages during development.

The passages come from 3 sources:

| Source | Graphic types |
| :- | :- |
| [Project Gutenberg](https://www.gutenberg.org/) ↗ | Illustrations from children's biographies |
| [Science Journal for Kids](https://www.sciencejournalforkids.org/) ↗ | Figures from science articles written for students |
| [Visualizing Energy](https://visualizingenergy.org/) ↗ | Interactive data graphics, captured as static screenshots |

Literacy experts from Student Achievement Partners (SAP) reviewed the evaluator's rating for each passage. They used a rubric based on the Use of Graphics dimension of SAP's Qualitative Text Complexity Rubric for Informational Text.

<Note>
  This dataset doesn't include the passages used to develop the evaluator. We
  plan to expand it with more graphic types and more Exceedingly Complex
  examples.
</Note>

## Our process

* After the evaluator was finalized, we used 20 new passages from the same 3 sources.
* Each graphic was prepared as an image file.
* We ran the evaluator 3 times on each passage, and kept its most common rating.
* 3 literacy experts from SAP independently reviewed each rating, noting whether they agree with it and scoring how sound its reasoning is on a scale of 1–5.
* We published only the passages where a majority of the experts agreed with the evaluator's rating. Of the 20 passages tested, 17 met that bar.

## Fields

<Note>
  Last updated October 14, 2026

  This list will be updated as the dataset expands.
</Note>

| Field | Definition |
| :- | :- |
| `id` | Identifier for the passage. Its prefix shows the source: `GUT` (Project Gutenberg), `SL` (Science Journal for Kids), or `VE` (Visualizing Energy). |
| `description` | Short description of the passage. |
| `input.text` | The passage text. |
| `input.grade_level` | Target grade for the rating, as a string (e.g., `"5"`). |
| `input.image_paths` | Paths to the passage's graphics, in reading order. |
| `input.figure_labels` | Labels for the graphics, in the same order as `input.image_paths`. |
| `expected.complexity_score` | The evaluator's rating, which a majority of the experts agreed with. The dataset includes `slightly_complex` (6 passages), `moderately_complex` (7 passages), and `very_complex` (4 passages). The evaluator can also return `exceedingly_complex`, and it returns `more_context_needed` when a graphic is absent, doesn't render, or is illegible. |

## Limitations

* The dataset is small (17 passages).
* It doesn't show cases where the experts disagreed with the evaluator's rating and reasoning.
* Each rating is the most common result of 3 runs. For 3 passages, the runs didn't all agree (they were never more than one level apart), so a single run may not match the published rating.
* The passages come from the same 3 sources used to develop the evaluator, so results may not generalize to other graphic types, like textbook diagrams, infographics, or comics.
* Only grades 4, 5, 7, 8, 10, and 12 are represented, even though the evaluator supports grades 3–12.
* There are no Exceedingly Complex examples.
* Each interactive graphic is represented by a single screenshot, so the dataset doesn't capture the demands of interacting with it.

## Related topics

<CardGroup cols={2}>
  <Card title="Datasets" icon="database" href="/evaluators/dataset/introduction">
    Understand what evaluator datasets are and how they are used.
  </Card>

  <Card title="Graphics evaluators" icon="chart-column" href="/evaluators/graphics-evaluators/introduction">
    Explore evaluators that assess the graphics in instructional content.
  </Card>

  <Card title="Core concepts" icon="lightbulb" href="/evaluators/understanding-evaluators/core-concepts">
    Learn how evaluator datasets inform accuracy and validation.
  </Card>
</CardGroup>