Next State Prediction on Atari Alien ALE (online)
79Chosen-Action NSP AccuracyMind-Studio
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Mind-StudioPlanner=DeepSeek-V4-Flash, Lookahead source=Mind-Studio (ours)2026.06 | 79 | 49 | |
| Mind-StudioPlanner=Claude-4-Sonnet, Lookahead source=Mind-Studio (ours)2026.06 | 55 | 38 | |
| Mind-StudioPlanner=Qwen3-Coder-480B, Lookahead source=Mind-Studio (ours)2026.06 | 50 | 38 | |
| PoE-WorldPlanner=Qwen3-Coder-480B, Lookahead source=PoE-World2026.06 | 48 | 28 | |
| Mind-StudioPlanner=Gemini-2.5-Flash, Lookahead source=Mind-Studio (ours)2026.06 | 26 | 24 | |
| Mind-StudioPlanner=Llama-4-Scout-17B, Lookahead source=Mind-Studio (ours)2026.06 | 24 | 20 | |
| LLM-as-WMPlanner=Qwen3-Coder-480B, Lookahead source=LLM-as-WM2026.06 | 0 | 8 |