News
Submission portal openedRegistration and benchmark submission are now available to participating teams.
WorldOracle team published a technical paper: WorldOracle: Benchmarking Action-Conditioned World Models against Real-World Executions
Five baseline models addedGE-Sim 2.0, Cosmos 3, DreamDojo, Ctrl-World and IRASim are now listed.
WorldOracle website launched
01 / Short-Horizon Prediction
WorldOracle-Core
The table reports Physical State Correctness (PSC), Key Interaction Area Fidelity (KIAF), Interaction Spatial Alignment (ISA), and Multi-View Action Coherence (MVAC, where supported). Single-View Score is the mean of PSC, KIAF and ISA. Multi-View Score adds MVAC with equal weight. A dash means unsupported or not evaluated.
Human Hand supports Single-View evaluation only. Multi-View Score and MVAC are not available.
AgiBot G2 · 30s · 8 Models
| Rank | Model | View | Contributor | SV Score | MV Score | PSC | KIAF | ISA | MVAC | Certification | Resources | Updated |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Loading published results… | ||||||||||||
No published model results for this platform and duration yet.
WorldOracle-Core, AgiBot G2Click any model to activate. (activated)
Matched Episode
Select a model to compare
Sampled episode from this model's evaluation materials.
Click any model row to show the example.
AgiBot G2 · 30s · All models shown equally.
02 / Action-Source Distribution
WorldOracle-Policy
The table reports Physical State Correctness (PSC), Key Interaction Area Fidelity (KIAF), Interaction Spatial Alignment (ISA), and Multi-View Action Coherence (MVAC, where supported). Single-View Score is the mean of PSC, KIAF and ISA. Multi-View Score adds MVAC with equal weight. A dash means unsupported or not evaluated.
AgiBot G2 · Policy Rollouts
| Rank | Model | View | Contributor | SV Score | MV Score | PSC | KIAF | ISA | MVAC | Certification | Resources | Updated |
|---|
WorldOracle-Policy, AgiBot G2Click any model to activate. (activated)
Matched Episode
Select a model to compare
Sampled episode from this model's evaluation materials.
Click any model row to show the example.
03 / Long-Horizon Prediction
WorldOracle-Long
The table reports Physical State Correctness (PSC), Key Interaction Area Fidelity (KIAF), Interaction Spatial Alignment (ISA), and Multi-View Action Coherence (MVAC, where supported). Single-View Score is the mean of PSC, KIAF and ISA. Multi-View Score adds MVAC with equal weight. A dash means unsupported or not evaluated.
Human Hand supports Single-View evaluation only. Multi-View Score and MVAC are not available.
AgiBot G2 · WorldOracle-Long
| Rank | Model | View | Contributor | SV Score | MV Score | PSC | KIAF | ISA | MVAC | Certification | Resources | Updated |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Loading leaderboard results… | ||||||||||||
No published model results for this platform and duration yet.
WorldOracle-Long, AgiBot G2Click any model to activate. (activated)
Matched Episode
Select a model to compare
Sampled episode from this model's evaluation materials.
Click any model row to show the example.
AgiBot G2 · WorldOracle-Long · All models shown equally.