Utility on G2U Utility
1Mean UtilityHuman
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| HumanParadigm=Control Group2025.11 | 1 | 2.9 | 61.6 | |
| PPOParadigm=Reinforcement Learning (RL) Group, Optimization Objective (@)=curiosity2025.11 | 0.609 | 6.9 | 29.3 | |
| PPOParadigm=Reinforcement Learning (RL) Group, Optimization Objective (@)=survival2025.11 | 0.588 | 7.3 | 30.5 | |
| PPOParadigm=Reinforcement Learning (RL) Group, Optimization Objective (@)=utility2025.11 | 0.534 | 8 | 25.6 | |
| DQNParadigm=Reinforcement Learning (RL) Group, Optimization Objective (@)=survival2025.11 | 0.497 | 10.8 | 15.2 | |
| Qwen3-VL-8BParadigm=Interactive Physical Reasoner Group, Interactive Physical Reasoner (IPR) Status=w/ IPR2025.11 | 0.493 | 8.5 | 22 | |
| DQNParadigm=Reinforcement Learning (RL) Group, Optimization Objective (@)=utility2025.11 | 0.445 | 10.8 | 17.1 | |
| DQNParadigm=Reinforcement Learning (RL) Group, Optimization Objective (@)=curiosity2025.11 | 0.424 | 10.9 | 15.9 | |
| GPT-5Paradigm=Multimodal Large Language Model (MLLM) Group, Optimization Objective (@)=utility2025.11 | 0.371 | 7.8 | 26.2 | |
| Qwen3-VL-8B-BCParadigm=Imitation Learning (IL) Group2025.11 | 0.368 | 9.6 | 12.8 | |
| GPT-4oParadigm=Multimodal Large Language Model (MLLM) Group, Optimization Objective (@)=utility2025.11 | 0.337 | 10 | 17.1 | |
| GPT-5Paradigm=Multimodal Large Language Model (MLLM) Group, Optimization Objective (@)=curiosity2025.11 | 0.333 | 9.8 | 16.5 | |
| GPT-4oParadigm=Multimodal Large Language Model (MLLM) Group, Optimization Objective (@)=survival2025.11 | 0.302 | 9.2 | 19.5 | |
| DreamerV3Paradigm=World Model Group, Optimization Objective (@)=survival2025.11 | 0.298 | 11.3 | 16.5 | |
| Qwen3-VL-30B-A3BParadigm=Multimodal Large Language Model (MLLM) Group, Optimization Objective (@)=survival2025.11 | 0.289 | 12 | 14 | |
| Qwen3-VL-30B-A3BParadigm=Multimodal Large Language Model (MLLM) Group, Optimization Objective (@)=utility2025.11 | 0.285 | 11.6 | 14.6 | |
| GPT-5Paradigm=Multimodal Large Language Model (MLLM) Group, Optimization Objective (@)=survival2025.11 | 0.263 | 8 | 23.8 | |
| ACT-BCParadigm=Imitation Learning (IL) Group2025.11 | 0.255 | 12 | 13.4 | |
| DreamerV3Paradigm=World Model Group, Optimization Objective (@)=curiosity2025.11 | 0.235 | 10 | 20.1 | |
| V-JEPA2Paradigm=World Model Group, Optimization Objective (@)=survival2025.11 | 0.191 | 13.9 | 16.5 | |
| GPT-4oParadigm=Multimodal Large Language Model (MLLM) Group, Optimization Objective (@)=curiosity2025.11 | 0.186 | 10.6 | 17.7 | |
| Qwen3-VL-8BParadigm=Interactive Physical Reasoner Group, Interactive Physical Reasoner (IPR) Status=w/o IPR2025.11 | 0.176 | 11.6 | 12.8 | |
| Qwen3-VL-30B-A3BParadigm=Multimodal Large Language Model (MLLM) Group, Optimization Objective (@)=curiosity2025.11 | 0.155 | 11.7 | 15.2 | |
| V-JEPA2Paradigm=World Model Group, Optimization Objective (@)=utility2025.11 | 0.152 | 14.1 | 11.6 | |
| V-JEPA2Paradigm=World Model Group, Optimization Objective (@)=curiosity2025.11 | 0.146 | 14 | 11.6 | |
| DreamerV3Paradigm=World Model Group, Optimization Objective (@)=utility2025.11 | 0.139 | 11.4 | 18.3 | |
| GenieReduxParadigm=World Model Group, Optimization Objective (@)=survival2025.11 | 0.128 | 12.7 | 12.8 | |
| GenieReduxParadigm=World Model Group, Optimization Objective (@)=utility2025.11 | 0.122 | 13.5 | 14.6 | |
| GenieReduxParadigm=World Model Group, Optimization Objective (@)=curiosity2025.11 | 0.1 | 12.8 | 12.8 | |
| RandomParadigm=Control Group2025.11 | 0 | 12.3 | 12.8 |