Long-Horizon Household Tasks on BEHAVIOR-1K
44.7FittingReflective Test-Time Planning
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Reflective Test-Time PlanningReflection-in-action (RIA)=true, Reflection-on-action (ROA)=true, Action model loss=true, Internal reflection model loss=true2026.02 | 44.7 | 32.4 | 31.7 | 25.8 | 33.65 | |
| Reflective Test-Time PlanningReflection-in-action (RIA)=false2026.02 | 33.5 | 5.88 | 3.17 | 3.23 | 11.45 | |
| Reflective Test-Time PlanningAction model loss=false2026.02 | 25.5 | 26.5 | 20.6 | 16.1 | 22.18 | |
| Reflective Test-Time PlanningInternal reflection model loss=false2026.02 | 12.8 | 8.82 | 17.5 | 9.68 | 12.2 | |
| Self-Refine2026.02 | 10.6 | 11.8 | 12.7 | 9.68 | 11.2 | |
| 3DLLM-Mem2026.02 | 10.6 | 14.7 | 9.52 | 9.68 | 11.13 | |
| Reflexion2026.02 | 8.51 | 8.82 | 15.9 | 6.45 | 9.92 | |
| Reflective Test-Time PlanningReflection-on-action (ROA)=false2026.02 | 6.38 | 11.8 | 19 | 12.9 | 12.52 | |
| DreamerV32026.02 | 4.26 | 11.8 | 11.1 | 12.9 | 10.02 | |
| ReflectVLM2026.02 | 2.12 | 5.88 | 14.3 | 6.45 | 7.19 | |
| PPO2026.02 | 0 | 2.94 | 7.94 | 3.23 | 3.53 | |
| Reflective Test-Time PlanningReflection-in-action (RIA)=false, Reflection-on-action (ROA)=false2026.02 | 0 | 17.6 | 11.1 | 12.9 | 10.4 |