Embodied Agent Planning on SafeAgentBench Safe Tasks
75.25Success RateNEXUS (Evolved)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| NEXUS (Evolved)Backbone=GPT-4, Knowledge Evolution Status=Evolved2026.05 | 75.25 | 1.34 | 16.19 | |
| LaMMA-PBackbone=GPT-4, Safety Prompting (Safety-LLM)=false, Execution Module=Identical2026.05 | 67.22 | 0 | 24.18 | |
| LaMMA-P (+ Safety LLM)Backbone=GPT-4, Safety Prompting (Safety-LLM)=true, Execution Module=Identical2026.05 | 64.55 | 2.68 | 30.82 | |
| ProgPromptBackbone=GPT-4, Safety Prompting (Safety-LLM)=false, Execution Module=Identical2026.05 | 63.21 | 0 | 28.71 | |
| ProgPrompt (+ Safety LLM)Backbone=GPT-4, Safety Prompting (Safety-LLM)=true, Execution Module=Identical2026.05 | 60.2 | 3.01 | 34.2 | |
| MLDTBackbone=GPT-4, Safety Prompting (Safety-LLM)=false, Execution Module=Identical2026.05 | 58.86 | 0 | 31.92 | |
| MLDT (+ Safety LLM)Backbone=GPT-4, Safety Prompting (Safety-LLM)=true, Execution Module=Identical2026.05 | 55.85 | 4.01 | 36.49 | |
| NEXUS (0-shot)Backbone=GPT-4, Knowledge Evolution Status=0-shot2026.05 | 55.18 | 3.58 | 28.89 | |
| ReActBackbone=GPT-4, Safety Prompting (Safety-LLM)=false, Execution Module=Identical2026.05 | 48.16 | 2.34 | 26.95 | |
| ReAct (+ Safety LLM)Backbone=GPT-4, Safety Prompting (Safety-LLM)=true, Execution Module=Identical2026.05 | 45.48 | 5.01 | 31.58 | |
| Lota-BenchBackbone=GPT-4, Safety Prompting (Safety-LLM)=false, Execution Module=Identical2026.05 | 38.13 | 0 | 20.78 | |
| Lota-Bench (+ Safety LLM)Backbone=GPT-4, Safety Prompting (Safety-LLM)=true, Execution Module=Identical2026.05 | 36.12 | 1.67 | 26.03 |