Songci Generation on Songci Evaluation Set
4.75FluencyERNIE 4.5 Turbo
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| ERNIE 4.5 TurboBest Prompt=completion2025.08 | 4.75 | 4.05 | 4.1 | 2.15 | 3.45 | 3.1 | 3.53 | |
| GPT-4oBest Prompt=zero-shot2025.08 | 4.72 | 4.26 | 4.35 | 2.1 | 3.6 | 3.2 | 3.5 | |
| DeepSeek-R1-0528-Qwen3-8BBest Prompt=zero-shot2025.08 | 4.1 | 3.8 | 3.6 | 1.8 | 2.9 | 2.7 | 2.85 | |
| Llama-3.3-70B-InstructBest Prompt=one-shot2025.08 | 4.05 | 3.6 | 3.4 | 1.75 | 2.8 | 2.55 | 2.7 | |
| Qwen3-32BBest Prompt=one-shot2025.08 | 4 | 3.4 | 3.2 | 1.65 | 2.6 | 2.35 | 2.5 | |
| Mistral-Small-Instruct-2409Best Prompt=one-shot2025.08 | 3.9 | 3.3 | 3.1 | 1.6 | 2.5 | 2.3 | 2.45 |