Layout prediction on HRS spatial
86.07AccuracyGPT-4 (reference)
Evaluation results
| Method | Links | |
|---|---|---|
| GPT-4 (reference)Model Size=>70B est.2024.03 | 86.07 | |
| Qwen3-8B + DivConModel Size=8B, Precision=INT42024.03 | 81.46 | |
| Qwen3-8BModel Size=8B, Precision=INT42024.03 | 59.39 | |
| Qwen3-4B + DivConModel Size=4B, Precision=FP162024.03 | 49.23 | |
| Mistral-7B + DivConModel Size=7B, Precision=INT42024.03 | 46.96 | |
| Phi-2 + DivConModel Size=2.7B, Precision=INT42024.03 | 41.51 | |
| TinyLlama + DivConModel Size=1.1B, Precision=INT42024.03 | 34.21 | |
| Mistral-7BModel Size=7B, Precision=FP162024.03 | 28.27 | |
| Qwen3-4BModel Size=4B, Precision=FP162024.03 | 24.43 | |
| Phi-2Model Size=2.7B, Precision=FP162024.03 | 23.56 | |
| TinyLlamaModel Size=1.1B, Precision=FP162024.03 | 17.6 |