Benchmarks

MobileGym

A browser-hosted mobile-app simulator with programmable state, parameterized tasks and deterministic state-checking judges for parallel GUI-agent evaluation and online reinforcement learning.

AI-assisted editorial review against a primary source. This is a catalog annotation, not an independent reproduction or endorsement.

Original source ↗Read as Markdown Record JSON

What is this resource?

Resource type: Benchmarks
Environment: web
Topics: evaluation, mobile-navigation, reinforcement-learning, sandbox
Record publication date: Not recorded
Evidence retrieved: 2026-10-17T11:18:19Z
Resource ID: github:purewhiter/mobilegym

What licenses are recorded?

Code, training data, weights, and documents can have different terms. A blank declaration means unknown, not unrestricted use.

Code: Apache-2.0 (Core source LICENSE; mobilegym-rl has a separate matching Apache-2.0 license and retains vendored notices.)
Data: CC-BY-NC-4.0 (Limited to maintainers’ dataset compilation/curation and authored synthetic content; third-party user content remains subject to platform terms and its authors’ rights.)
Document: Not recorded
Weights: Not recorded

License review notes

Code: Core source LICENSE; mobilegym-rl has a separate matching Apache-2.0 license and retains vendored notices. Source 1
Data: Limited to maintainers’ dataset compilation/curation and authored synthetic content; third-party user content remains subject to platform terms and its authors’ rights. Source 1

Evaluation protocol

Focus: mobile GUI task completion in simulation, state verification and parallel training
Setting: interactive
Observations: simulated mobile screenshots, structured simulator state for judges
Actions / outputs: tap, type, swipe, back, home, drag, complete
Evaluation: Deterministic task judges inspect state changes; reported metrics distinguish success, progress, false completion and unexpected side effects.

Scale and counting units

416 task templates — publisher benchmark overview source
28 simulated apps — publisher benchmark overview source
160 task identifiers — bench_env/splits/train.txt (computed) source
256 task identifiers — bench_env/splits/test.txt (computed) source

The split-ID inventory is not a count of sampled trajectories, live Android apps or real user accounts. Extra runtime/data dependencies and third-party content terms remain distinct.

Compare source-reviewed profiles

Recorded specifications

availability: Public simulator, task/judge and online-RL source, plus official release metadata for the companion data archive.
release tag: v0.1.0
repository revision: 5fbd3f34be0e9780ae4b1e9d5089963ddbd05d38

Explore related topics

Evaluation Mobile navigation Reinforcement learning Sandboxes

Scope and limitations

The inspected code release is v0.1.0, with a separately listed data-v0.1.0 archive. The interaction target is a browser-hosted Android-like simulation; real-device transfer results are separate publisher-reported experiments. Source split manifests contain 160 training IDs and 256 test IDs with no duplicates or overlap. Platform/training code declares Apache-2.0. LICENSE-DATA applies CC-BY-NC-4.0 only to maintainer-created compilation, curation and synthetic content, and explicitly leaves third-party UGC/API terms and author rights separate. No runtime, archive or model was executed/downloaded.

Related resources

documented by: MobileGym: A Verifiable and Highly Parallel Simulation Platform for Mobile GUI Agent Research

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 2026-10-17T11:18:19Z. Supports: name, kind, summary, environments, topics, scope, notes, relations, details.benchmark_protocol.
Content fingerprint
815294dd2aa4bf8264978c3bfedee85a3a2c835c511860c732314a4c282a974a
editorial primary review
Retrieved 2026-10-17T11:18:18Z. Supports: identity.
Content fingerprint
48df04a2fbf479f956d5c7f22d7006c22f591b8857cab8a1d64c196349045495
editorial primary review
Retrieved 2026-10-17T11:22:56Z. Supports: release_status, details.repository_revision.
Content fingerprint
68f1fa16c0999079bccc3aa1b6c3c22b72b570a5b971837c5026ea5408853375
editorial primary review
Retrieved 2026-10-17T11:23:08Z. Supports: release_status, details.release_tag.
Content fingerprint
cb1057b0a0a2ef7ab781414c168bb41f0f3da88afc042ce266d6063493561bde
editorial primary review
Retrieved 2026-10-17T11:23:09Z. Supports: release_status.
Content fingerprint
ad0c86a6cdde2caba8bd1f98b121803817d93718319a863b0b833da2760c23aa
editorial primary review
Retrieved 2026-10-17T11:22:59Z. Supports: licenses.code.
Content fingerprint
7618cd7de039c264759ca8ca0b9130c6de63a4536dcf436baedd4cfbf639073e
editorial primary review
Retrieved 2026-10-17T11:27:50Z. Supports: licenses.code.
Content fingerprint
58d1e17ffe5109a7ae296caafcadfdbe6a7d176f0bc4ab01e12a689b0499d8bd
editorial primary review
Retrieved 2026-10-17T11:22:59Z. Supports: licenses.data, notes.
Content fingerprint
e222fa518c5003d425ee9b371e987c95ba10e52b54fea078571254ae6a8eaf23
editorial primary review
Retrieved 2026-10-17T11:27:49Z. Supports: details.benchmark_protocol.
Content fingerprint
0ecef460fecfd8aa91dcbb2cd4cca17b033f620f85a50d7f0b598ed19a27c0eb
editorial primary review
Retrieved 2026-10-17T11:27:48Z. Supports: details.benchmark_protocol.
Content fingerprint
a3d485015d480f4f1ebbd12dbbf02c25c0aa667d1afcf5f6fa35f0263a82c567

Suggest a correction

Open a source-backed correction. Include this record ID and the original source. Read our editorial policy and coverage boundaries.