Furniture Layout Generation on CubiCasa5K (out-of-distribution)
-15.31Mean ScoreQwen3.5-9B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen3.5-9BRoom=Kitchen, Mode=zero-shot, K=10, Protocol=Rule-scorer2026.06 | -15.31 | 6.6 | -1.07 | |
| Qwen3.5-9BRoom=Living room, Mode=zero-shot, K=10, Protocol=Rule-scorer2026.06 | -8.05 | 4.88 | 2.15 | |
| Qwen3.5-9BRoom=Overall, Mode=zero-shot, K=10, Protocol=Rule-scorer2026.06 | -8.02 | 5.38 | 2.04 | |
| Qwen3.5-9BRoom=Bedroom, Mode=zero-shot, K=10, Protocol=Rule-scorer2026.06 | -6.5 | 5.12 | 2.54 | |
| GLM-5V-TurboRoom=Kitchen, Mode=zero-shot, K=10, Protocol=Rule-scorer2026.06 | -5.51 | 6.01 | -1.17 | |
| Kimi K2.5Room=Kitchen, Mode=zero-shot, K=10, Protocol=Rule-scorer2026.06 | -3.25 | 7.17 | 0.76 | |
| Qwen3.5-9BRoom=Bathroom, Mode=zero-shot, K=10, Protocol=Rule-scorer2026.06 | -2.21 | 4.64 | 4.53 | |
| Architect-AntRoom=Kitchen, Mode=SFT + synthetic-pair DPO, K=10, Protocol=Rule-scorer2026.06 | -1.23 | 4.69 | 5.81 | |
| SFT (AntPlan-270)Room=Kitchen, Mode=supervised fine-tuning, K=10, Protocol=Rule-scorer2026.06 | -1.22 | 4.71 | 5.46 | |
| Architect-AntRoom=Living room, Mode=SFT + synthetic-pair DPO, K=10, Protocol=Rule-scorer2026.06 | -0.46 | 4.92 | 7.06 | |
| GLM-5V-TurboRoom=Overall, Mode=zero-shot, K=10, Protocol=Rule-scorer2026.06 | -0.43 | 4.95 | 2.75 | |
| GLM-5V-TurboRoom=Living room, Mode=zero-shot, K=10, Protocol=Rule-scorer2026.06 | -0.42 | 5.14 | 3.16 | |
| SFT (AntPlan-270)Room=Living room, Mode=supervised fine-tuning, K=10, Protocol=Rule-scorer2026.06 | -0.34 | 4.75 | 7.39 | |
| Kimi K2.5Room=Living room, Mode=zero-shot, K=10, Protocol=Rule-scorer2026.06 | 0.14 | 5.15 | 3.72 | |
| GLM-5V-TurboRoom=Bedroom, Mode=zero-shot, K=10, Protocol=Rule-scorer2026.06 | 0.77 | 4.29 | 3.4 | |
| SFT (AntPlan-270)Room=Overall, Mode=supervised fine-tuning, K=10, Protocol=Rule-scorer2026.06 | 1.02 | 4.26 | 7.27 | |
| Kimi K2.5Room=Overall, Mode=zero-shot, K=10, Protocol=Rule-scorer2026.06 | 1.26 | 5.27 | 4.26 | |
| Architect-AntRoom=Overall, Mode=SFT + synthetic-pair DPO, K=10, Protocol=Rule-scorer2026.06 | 1.42 | 4.21 | 7.34 | |
| SFT (AntPlan-270)Room=Bathroom, Mode=supervised fine-tuning, K=10, Protocol=Rule-scorer2026.06 | 2.67 | 3.7 | 8.1 | |
| SFT (AntPlan-270)Room=Bedroom, Mode=supervised fine-tuning, K=10, Protocol=Rule-scorer2026.06 | 2.95 | 3.78 | 8.14 | |
| Architect-AntRoom=Bedroom, Mode=SFT + synthetic-pair DPO, K=10, Protocol=Rule-scorer2026.06 | 3.04 | 3.58 | 8.21 | |
| Architect-AntRoom=Bathroom, Mode=SFT + synthetic-pair DPO, K=10, Protocol=Rule-scorer2026.06 | 3.32 | 3.5 | 8.26 | |
| GLM-5V-TurboRoom=Bathroom, Mode=zero-shot, K=10, Protocol=Rule-scorer2026.06 | 3.46 | 4.04 | 5.62 | |
| Kimi K2.5Room=Bedroom, Mode=zero-shot, K=10, Protocol=Rule-scorer2026.06 | 4.03 | 3.46 | 6.62 | |
| Kimi K2.5Room=Bathroom, Mode=zero-shot, K=10, Protocol=Rule-scorer2026.06 | 4.11 | 4.53 | 5.92 |