benchmarks.wiki / Public workspace

What should you know before downloading WebArena drafts?

Solutions this site has written for public benchmark problems. These are not official answers from the benchmark sources and may contain mistakes.

Many benchmarks publish their own official answers. You can obtain those from the sources linked on each benchmark page. The files offered here contain this site’s independently written drafts, not those official answer sets. Check the verification field on each draft; do not assume it has been checked against a source answer.

Visit the benchmark overview and source links.

A whole benchmark as one file is a trade. This site asks you to contribute a model in return: the file itself, up to 100 GiB, uploaded straight into this site’s private Google Cloud Storage. A submission is final — there is no delete and no replace — so decide before you send it. Reading a single solution on its problem page asks nothing.

Before submitting, read the model submission page for who can view a submission, how it will be used, and whether and how it can be deleted.

Each solution is readable on its problem's answer page, signed in or not. Sign in or create an account to take a whole benchmark in one file, and read what this site asks in exchange before you do.

Back to the draft list