---
title: "Public benchmark and software-paper reproduction"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Public benchmark and software-paper reproduction}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(collapse = TRUE, comment = "#>")
```

# Bundled multimodal study

The package includes a compact, fully synthetic and openly redistributable benchmark containing participant and item metadata, binary responses, response times, recordings, gaze samples, events, AOIs, pupil trajectories, quality indicators, and provenance.

```{r}
library(eyeprocess)
study <- eyeprocess_benchmark_study()
study
```

The benchmark tests contracts and reproducibility. It is not real participant data, vendor compatibility evidence, or scientific validation of advanced models.

# Integrity and expected outputs

```{r}
validation <- validate_benchmark_study(study)
validation$valid

reproduction <- run_benchmark_reproduction(study)
reproduction$comparison
```

All source files are fingerprinted. Relational checks ensure that participants, items, trials, events, gaze, and pupil records agree. Expected scalar outputs use explicit numerical tolerances.

# Data dictionary and reproduction directory

```{r, eval=FALSE}
write_benchmark_data_dictionary(study, "benchmark-data-dictionary.md")
write_software_paper_reproduction("software-paper-reproduction", study)
```

The reproduction scaffold contains copied benchmark data, an executable R script, expected outputs, a README, and a file/session manifest.

# Release audit

```{r}
audit_benchmark_release(study)$ready
```

A public empirical benchmark can later be added under an independently reviewed licence. It must remain distinct from this synthetic test asset.