Proactive Assistance on Gridworld
24.5SpeedupMindZero
Evaluation results
| Method | Links | |
|---|---|---|
| MindZeroBackbone=Qwen3-VL-8B, TFLOPs=291.82026.05 | 24.5 | |
| MindZeroBackbone=Qwen3-VL-4B, TFLOPs=161.42026.05 | 23 | |
| Qwen3-VL-4BTFLOPs=151.72026.05 | 1.4 | |
| Qwen3-VL-235B-A22BTFLOPs=808.62026.05 | 1 | |
| Random Goal2026.05 | 0 | |
| GPT-5.2TFLOPs=Proprietary2026.05 | 0 | |
| Gemini-3-FlashTFLOPs=Proprietary2026.05 | 0 | |
| Qwen3-VL-8BTFLOPs=295.22026.05 | -0.1 |