Attribute Value Extraction on Amazon Reviews 2023
88.4F1LLMQwen3-32B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Qwen3-32BDecoding Strategy=AR, Model Type=Fine-tuned model2026.04 | 88.4 | 7.745 | |
| Qwen3-32BDecoding Strategy=HPD, Model Type=Fine-tuned model2026.04 | 88.4 | 0.898 | |
| Phi4-14BDecoding Strategy=HPD, Model Type=Fine-tuned model2026.04 | 88.3 | 0.248 | |
| Qwen3-8BDecoding Strategy=HPD, Model Type=Fine-tuned model2026.04 | 88.1 | 0.167 | |
| Phi4-14BDecoding Strategy=AR, Model Type=Fine-tuned model2026.04 | 87.6 | 3.42 | |
| Qwen3-8BDecoding Strategy=AR, Model Type=Fine-tuned model2026.04 | 87.4 | 1.362 | |
| GPT-4(.1)Prompting Setting=0-shot, Model Type=Base model2026.04 | 87.1 | 1.14 | |
| Qwen3-4BDecoding Strategy=HPD, Model Type=Fine-tuned model2026.04 | 87 | 0.133 | |
| Qwen3-1.7BDecoding Strategy=HPD, Model Type=Fine-tuned model2026.04 | 87 | 0.095 | |
| Qwen3-4BDecoding Strategy=AR, Model Type=Fine-tuned model2026.04 | 86.1 | 1.2 | |
| Qwen3-1.7BDecoding Strategy=AR, Model Type=Fine-tuned model2026.04 | 86.1 | 1.11 | |
| Phi4-14BDecoding Strategy=AR, Model Type=Base model2026.04 | 83.9 | 2.675 | |
| Phi4-14BDecoding Strategy=HPD, Model Type=Base model2026.04 | 81.4 | 0.216 | |
| Qwen3-32BDecoding Strategy=HPD, Model Type=Base model2026.04 | 79.7 | 0.805 | |
| Qwen3-32BDecoding Strategy=AR, Model Type=Base model2026.04 | 79.1 | 5.347 | |
| Qwen3-8BDecoding Strategy=HPD, Model Type=Base model2026.04 | 75.1 | 0.132 | |
| Qwen3-4BDecoding Strategy=HPD, Model Type=Base model2026.04 | 73.2 | 0.127 | |
| Qwen3-4BDecoding Strategy=AR, Model Type=Base model2026.04 | 67.1 | 1.12 | |
| Qwen3-8BDecoding Strategy=AR, Model Type=Base model2026.04 | 64.6 | 1.142 | |
| Qwen3-1.7BDecoding Strategy=HPD, Model Type=Base model2026.04 | 52.9 | 0.09 | |
| Qwen3-1.7BDecoding Strategy=AR, Model Type=Base model2026.04 | 51 | 0.887 |