{"$schema":"https://botlandscape.com/claims.schema.json","schema_version":1,"projection":"filtered-compare-index","source_schema_version":3,"compare_index_url":"https://botlandscape.com/compare-index.json","entity_id":"benchmark--robotwin-2","convention":{"class":"exact data-class","metric_id":"exact row id","state":"reported or missing","evidence_grade":"manual, filing, research, press, or analyst"},"counts":{"tables":1,"rows":16,"claims":16,"sources":2},"tables":[{"key":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval","page_path":"topics/benchmarks.html","id":"vla-manipulation--vla-manipulation-eval","class":"vla-manipulation-eval","label":"Robotics benchmarks · Manipulation evals","caption":"Sources: VLABench paper ; VLABench project ; RoboTwin 2.0 paper ; RoboTwin 2.0 project . Identity and protocol only; method scores remain in the papers and official leaderboards.","subjects":[{"id":"robotwin-2","label":"RoboTwin 2.0"}],"rows":[{"id":"vla-manipulation--kind","metric":"Kind","url":"https://botlandscape.com/benchmarks#vla-manipulation--kind","cells":[{"id":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval::vla-manipulation--kind::robotwin-2","column":2,"subject_ids":["robotwin-2"],"state":"reported","text":"scalable bimanual data generator and policy-generalization benchmark","evidence":["benchmarks-r34"],"links":[]}]},{"id":"vla-manipulation--first-public-date","metric":"First public date","url":"https://botlandscape.com/benchmarks#vla-manipulation--first-public-date","cells":[{"id":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval::vla-manipulation--first-public-date::robotwin-2","column":2,"subject_ids":["robotwin-2"],"state":"reported","text":"arXiv 2025-06-22","evidence":["benchmarks-r34"],"links":[]}]},{"id":"vla-manipulation--paper","metric":"Paper","url":"https://botlandscape.com/benchmarks#vla-manipulation--paper","cells":[{"id":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval::vla-manipulation--paper::robotwin-2","column":2,"subject_ids":["robotwin-2"],"state":"reported","text":"arXiv:2506.18088","evidence":["benchmarks-r34"],"links":[]}]},{"id":"vla-manipulation--task-structure","metric":"Task structure","url":"https://botlandscape.com/benchmarks#vla-manipulation--task-structure","cells":[{"id":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval::vla-manipulation--task-structure::robotwin-2","column":2,"subject_ids":["robotwin-2"],"state":"reported","text":"50 dual-arm tasks in the paper; the project describes over 50 supported tasks","evidence":["benchmarks-r34","benchmarks-r35"],"links":[]}]},{"id":"vla-manipulation--asset-library","metric":"Asset library","url":"https://botlandscape.com/benchmarks#vla-manipulation--asset-library","cells":[{"id":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval::vla-manipulation--asset-library::robotwin-2","column":2,"subject_ids":["robotwin-2"],"state":"reported","text":"RoboTwin-OD: 731 objects across 147 categories","evidence":["benchmarks-r34"],"links":[]}]},{"id":"vla-manipulation--released-trajectories","metric":"Released trajectories","url":"https://botlandscape.com/benchmarks#vla-manipulation--released-trajectories","cells":[{"id":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval::vla-manipulation--released-trajectories::robotwin-2","column":2,"subject_ids":["robotwin-2"],"state":"reported","text":"100,000+ pre-collected expert trajectories across 50 tasks and 5 embodiments","evidence":["benchmarks-r34"],"links":[]}]},{"id":"vla-manipulation--evaluation-scope","metric":"Evaluation scope","url":"https://botlandscape.com/benchmarks#vla-manipulation--evaluation-scope","cells":[{"id":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval::vla-manipulation--evaluation-scope::robotwin-2","column":2,"subject_ids":["robotwin-2"],"state":"reported","text":"5 randomization axes: clutter, lighting, background, tabletop height, and language; clean versus randomized policy evaluation","evidence":["benchmarks-r34"],"links":[]}]},{"id":"vla-manipulation--medium","metric":"Medium","url":"https://botlandscape.com/benchmarks#vla-manipulation--medium","cells":[{"id":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval::vla-manipulation--medium::robotwin-2","column":2,"subject_ids":["robotwin-2"],"state":"reported","text":"simulation benchmark and data generator with a separate sim-to-real study","evidence":["benchmarks-r34"],"links":[]}]},{"id":"vla-manipulation--embodiment-in-the-source","metric":"Embodiment in the source","url":"https://botlandscape.com/benchmarks#vla-manipulation--embodiment-in-the-source","cells":[{"id":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval::vla-manipulation--embodiment-in-the-source::robotwin-2","column":2,"subject_ids":["robotwin-2"],"state":"reported","text":"dual-arm pairings of Aloha-AgileX, ARX-X5, Piper, Franka, and UR5 manipulators","evidence":["benchmarks-r34"],"links":[]}]},{"id":"vla-manipulation--engine-renderer","metric":"Engine / renderer","url":"https://botlandscape.com/benchmarks#vla-manipulation--engine-renderer","cells":[{"id":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval::vla-manipulation--engine-renderer::robotwin-2","column":2,"subject_ids":["robotwin-2"],"state":"missing","text":"—","evidence":[],"links":[]}]},{"id":"vla-manipulation--primary-metric-family","metric":"Primary metric family","url":"https://botlandscape.com/benchmarks#vla-manipulation--primary-metric-family","cells":[{"id":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval::vla-manipulation--primary-metric-family::robotwin-2","column":2,"subject_ids":["robotwin-2"],"state":"reported","text":"task success under clean and hard randomized conditions on the full 50-task protocol","evidence":["benchmarks-r34"],"links":[]}]},{"id":"vla-manipulation--language-conditioned","metric":"Language-conditioned","url":"https://botlandscape.com/benchmarks#vla-manipulation--language-conditioned","cells":[{"id":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval::vla-manipulation--language-conditioned::robotwin-2","column":2,"subject_ids":["robotwin-2"],"state":"reported","text":"yes — trajectory-level language is one randomization axis","evidence":["benchmarks-r34"],"links":[]}]},{"id":"vla-manipulation--public-code","metric":"Public code","url":"https://botlandscape.com/benchmarks#vla-manipulation--public-code","cells":[{"id":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval::vla-manipulation--public-code::robotwin-2","column":2,"subject_ids":["robotwin-2"],"state":"reported","text":"yes — MIT-licensed official repository","evidence":["benchmarks-r35"],"links":[]}]},{"id":"vla-manipulation--public-data","metric":"Public data","url":"https://botlandscape.com/benchmarks#vla-manipulation--public-data","cells":[{"id":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval::vla-manipulation--public-data::robotwin-2","column":2,"subject_ids":["robotwin-2"],"state":"reported","text":"yes — object library and pre-collected trajectories are linked from the official project","evidence":["benchmarks-r35"],"links":[]}]},{"id":"vla-manipulation--public-leaderboard","metric":"Public leaderboard","url":"https://botlandscape.com/benchmarks#vla-manipulation--public-leaderboard","cells":[{"id":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval::vla-manipulation--public-leaderboard::robotwin-2","column":2,"subject_ids":["robotwin-2"],"state":"reported","text":"yes — official leaderboard linked from the project","evidence":["benchmarks-r35"],"links":[]}]},{"id":"vla-manipulation--claim-boundary","metric":"Claim boundary","url":"https://botlandscape.com/benchmarks#vla-manipulation--claim-boundary","cells":[{"id":"topics/benchmarks.html::vla-manipulation--vla-manipulation-eval::vla-manipulation--claim-boundary::robotwin-2","column":2,"subject_ids":["robotwin-2"],"state":"reported","text":"50 benchmark tasks are distinct from the 10-task code-generation study and 4-task real-world study; those method scores stay in the paper","evidence":["benchmarks-r34"],"links":[]}]}]}],"sources":[{"id":"benchmarks-r34","marker":"R34","grade":"research","citation":"RoboTwin 2.0 — “RoboTwin 2.0: A Scalable Data Generator and Benchmark with Strong Domain Randomization for Robust Bimanual Robotic Manipulation”","access_status":"direct","registry_url":"https://botlandscape.com/sources.html#benchmarks-r34","direct_url":"https://arxiv.org/abs/2506.18088"},{"id":"benchmarks-r35","marker":"R35","grade":"research","citation":"RoboTwin 2.0 — official project site with linked code, data, documentation, and leaderboard","access_status":"direct","registry_url":"https://botlandscape.com/sources.html#benchmarks-r35","direct_url":"https://robotwin-platform.github.io/"}]}