MockWeb-Bench

MockWeb-Bench provides website reconstruction tasks with reference websites and evaluation material. This release contains 50 complete site packages, version v3_relicensed_260825-assets.

Each site.tar.gz expands to a directory named after its site and includes the reference website under site/, task.json, site_meta.json, evaluation material under evaluation/, and the supplied licensing files. The archives are preserved byte-for-byte from this release.

The compressed download is 49.71 GB (49,712,660,947 bytes). Files after unpacking total approximately 54.25 GB, excluding filesystem overhead. Allow space for both the downloaded archives and the extracted files.

Download

Install the Hugging Face Hub client and download the release:

python -m pip install -U huggingface_hub
hf download submission11898/MockWeb-Bench --repo-type dataset --revision v3_relicensed_260825-assets --local-dir mockweb-data

To download one site from Python:

from huggingface_hub import hf_hub_download

hf_hub_download(
    repo_id="submission11898/MockWeb-Bench",
    repo_type="dataset",
    revision="v3_relicensed_260825-assets",
    filename="v3_relicensed_260825-assets/glyphel.example/site.tar.gz",
    local_dir="mockweb-data",
)

Verify and extract

The manifest records every archive's size, SHA-256, site identifier, archive root and licensing-file paths. Verify a complete download with:

cd mockweb-data
sha256sum -c SHA256SUMS

For a run on one site, extract that site's archive into a new dataset directory:

mkdir dataset-glyphel
tar -xzf v3_relicensed_260825-assets/glyphel.example/site.tar.gz -C dataset-glyphel

For all sites, keep a separate dataset directory per site. This accommodates the standalone runner's limit of 100,000 files in the directory it inventories; the complete collection contains more files than that:

mkdir datasets
for archive in v3_relicensed_260825-assets/*/site.tar.gz; do
  site=$(basename "$(dirname "$archive")")
  mkdir "datasets/$site"
  tar -xzf "$archive" -C "datasets/$site"
done

Each resulting datasets/<site>/ directory can be passed to --dataset with its matching --site <site> argument.

The resulting layout is:

dataset-glyphel/
  glyphel.example/
    LICENSE
    task.json
    site_meta.json
    site/
    evaluation/
  ...

Pass this extracted directory to the MockWeb-Bench runner using --dataset, and select a site using its directory name. These per-site archives should first be extracted; the runner's single-archive input expects a different, manifest-wrapped dataset format.

For example, from the runner repository, after setting your model configuration:

mockweb run --backend docker \
  --dataset /path/to/mockweb-data/dataset-glyphel \
  --site glyphel.example \
  --config /path/to/config.json \
  --output /path/to/results

Evaluation material belongs to the evaluator. Use the runner's isolation and staging workflow to keep it outside the reconstruction agent's workspace.

Sites

Site	Compressed MB	File
ablira.example	481.1	archive
aperlio.example	6980.3	archive
astrovia.example	263.2	archive
borvia.example	1163.9	archive
bramble.example	88.5	archive
breezel.example	1069.9	archive
buildara.example	358.6	archive
cascora.example	1320.4	archive
climatewatch.org	219.3	archive
corravale.example	131.4	archive
cresvia.example	582.4	archive
detox.github.io	469.6	archive
fluvica.example	430.2	archive
formora.example	1388.1	archive
fynora.example	770.4	archive
glyphel.example	76.5	archive
gradyn.example	576.3	archive
haulvio.example	489.9	archive
inclura.example	885.1	archive
kartova.example	522.9	archive
keycloak.org	269.5	archive
klaravo.example	539.4	archive
klavrium.example	933.7	archive
mendova.example	472.6	archive
nimbric.example	270.1	archive
obseron.example	5210.7	archive
postgresql.org	461.6	archive
praxide.example	115.3	archive
praxora.example	1296.3	archive
quorvex.example	385.7	archive
rankspire.example	3702.5	archive
relvyn.example	563.3	archive
rituvia.example	622.3	archive
rust-lang.org	91.6	archive
sentorra.example	1728.2	archive
sonelio.example	980.5	archive
sonivex.example	915.0	archive
streamora.example	755.5	archive
talvio.example	1626.6	archive
terravyn.example	726.2	archive
vaultera.example	1133.9	archive
vellora.example	853.4	archive
veranthe.example	419.3	archive
verendale.example	631.5	archive
verentia.example	2282.3	archive
veridara.example	2629.7	archive
verrindale.example	390.7	archive
vizmara.example	588.7	archive
weavora.example	193.8	archive
webpack.js.org	654.9	archive

Licensing

The original per-site LICENSE/NOTICE files are preserved inside the archives and copied to LICENSES for convenient review. Licensing varies by site and component; third-party asset notices continue to apply. See the license index for the declarations supplied with each package.

Dataset issues

For a missing or damaged file, report the site identifier, relative file path and archive SHA-256 in this repository's discussions.