| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Explainable AI | CRAFTER-XAI-Bench (test) | Faithfulness0.54 | 25 | |
| Sequential environment decision making | Crafter BALROG protocol | Peak Task Score (%)57.3 | 20 | |
| AI-generated video detection | Crafter | Recall100 | 10 | |
| Embodied Task Completion | Crafter 5×5 Mini | Cumulative Return289 | 7 | |
| Question Answering | Crafter four-actor rollout robustness | Average Exact Match (EM)59.3 | 7 | |
| Sequential Decision Making | CrafterMini 5x5 | Cumulative Return150.3 | 5 | |
| Environment Task Completion | CrafterMini 5x5 | Normalized Success Rate100 | 5 | |
| Raster-to-SVG conversion | 80 CRAFTER outputs (test) | Position Score8.1 | 5 | |
| World Modeling | Crafter-OO | Rank @ 118.7 | 5 | |
| Reinforcement Learning | Crafter 1M steps | Score12.1 | 5 | |
| Latent Action Policy Learning | Crafter | Wood Harvest Rate86 | 4 | |
| Reinforcement Learning | Crafter | Score16.2 | 4 | |
| World Model Prediction | Crafter | Avg Imagination MSE5.07 | 2 |