WONDERBREAD
An offline benchmark of workflow documentation, quality assessment and knowledge transfer using human web-navigation demonstrations, action traces and written procedures.
AI-assisted editorial review against a primary source. This is a catalog annotation, not an independent reproduction or endorsement.
What is this resource?
- Resource type
- Benchmarks
- Environment
- web
- Topics
- evaluation, workflow, video, screen-understanding
- Record publication date
- Not recorded
- Evidence retrieved
- 2026-10-17T09:12:11Z
- Resource ID
- github:hazyresearch/wonderbread
What licenses are recorded?
Code, training data, weights, and documents can have different terms. A blank declaration means unknown, not unrestricted use.
- Code
- Not recorded
- Data
- cc-by-4.0 (The linked Zenodo dataset release declares CC-BY-4.0; this does not assign a code license to the repository.)
- Document
- Not recorded
- Weights
- Not recorded
License review notes
- Data: The linked Zenodo dataset release declares CC-BY-4.0; this does not assign a code license to the repository. Source 1
Evaluation protocol
- Focus
- workflow documentation, workflow quality assessment, workflow knowledge transfer
- Setting
- offline
- Observations
- screen recordings, action traces, screenshots, written standard operating procedures
- Actions / outputs
- generate procedures, segment demonstrations, answer workflow questions, validate and rank procedures
- Evaluation
- Task-specific scripts evaluate six workflow-understanding task types; model outputs are not treated as successful interactive execution.
Scale and counting units
- 6 task types — WONDERBREAD BPM benchmark source
- 2,928 human demonstrations — full dataset release 14094162 source
- 598 web navigation workflows — full dataset release 14094162 source
- 724 human demonstrations — gold subset used for the original paper evaluations source
Counts are publisher-reported. The gold subset covers 162 workflows and differs from the full collection; scores require the exact task and subset.
Compare source-reviewed profiles
Recorded specifications
- availability
- Public benchmark code and public Zenodo dataset release metadata.
- dataset release
- 10.5281/zenodo.14094162
- repository revision
- ed052c67aeada04167cdfe92ff8de454aa94627a
Explore related topics
Scope and limitations
The repository contains evaluation and collection code; the linked Zenodo release exposes the full, gold and debug demonstration archives and metadata. This measures workflow understanding rather than interactive agent task completion. Zenodo declares CC-BY-4.0 for that dataset; no repository-level code license was found. Archives and human demonstration records were not downloaded.
Related resources
- documented by: WONDERBREAD: A Benchmark for Evaluating Multimodal Foundation Models on Business Process Management Tasks
Primary evidence and verification
Source retrieval is separate from experiment reproduction. Use the original publication or publisher page when citing claims about the resource.
- editorial primary review
Retrieved 2026-10-17T09:12:11Z. Supports: name, kind, summary, environments, topics, scope, notes, relations, details.
Content fingerprint
1e4308d59a22b39543c28b745a58d3d7c47be1ab9df9a4c5c0fe868a810350b2 - editorial primary review
Retrieved 2026-10-17T09:12:11Z. Supports: identity.
Content fingerprint
5ba27e24455b36b52bf65b3c9f589e760f97a576bb01d622d89836c5bb53b38c - editorial primary review
Retrieved 2026-10-17T09:14:04Z. Supports: release_status, licenses.code.
Content fingerprint
accb518a5362948cf96e9f0b684fb6fcc9b26e43a691016ae70a3283c0e1e2c2 - editorial primary review
Retrieved 2026-10-17T09:15:34Z. Supports: release_status, licenses.data, details.benchmark_protocol.
Content fingerprint
68b4b952c1ab90484d457b392de1d0c95a24f6cc1674c82125a6bd26f1cb1016
Suggest a correction
Open a source-backed correction. Include this record ID and the original source. Read our editorial policy and coverage boundaries.