COSMOS
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Next-Step-Prediction Style Planning | Cosmos Reason | Performance66.82 | 16 | |
| Embodied Planning & Correction | Cosmos | Accuracy69.3 | 14 | |
| Image-to-Video | Cosmos HUE | Human Evaluation Score94.4 | 11 | |
| Text-to-Video | Cosmos HUE | Human Evaluation Score93.6 | 11 | |
| Planning | Cosmos | Score66.36 | 10 | |
| Log-likelihood Estimation | COSMOS 2020 (test) | Mean0 | 9 | |
| Embodied Reasoning | Cosmos R1 | Score81 | 9 | |
| Multi-modal Forgery Detection | COSMOS | ACC53.78 | 5 | |
| Cross-modal consistency verification | COSMOS | Accuracy56.85 | 5 | |
| Video Generation Inference | Cosmos3-Super 64B | Latency (s)43.9 | 3 |
Showing 10 of 10 rows