Plain Summarization on BillSum
46.7ROUGE-1GPT-4o
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| GPT-4oRevision Stage=Initial Summary2026.04 | 46.7 | 87.07 | 17.29 | 13.11 | 16.56 | 60.91 | 34.41 | |
| Llama-3.1-8B-InstructRevision Stage=Initial Summary2026.04 | 45.81 | 86.83 | 18.04 | 12.77 | 15.13 | 58.38 | 33.48 | |
| GPT-4oRevision Stage=Round 1, Revision Method=NRLB2026.04 | 43.28 | 86.43 | 12.64 | 12.06 | 14.07 | 68.2 | 39.41 | |
| Llama-3.1-8B-InstructRevision Stage=Round 1, Revision Method=NRLB2026.04 | 42.66 | 86.2 | 15.66 | 11.88 | 13.59 | 62.58 | 34.51 | |
| Qwen3-8BRevision Stage=Initial Summary2026.04 | 41.08 | 85.61 | 18.07 | 14.14 | 17.38 | 56.29 | 30.69 | |
| GPT-4oRevision Stage=Round 2, Revision Method=NRLB2026.04 | 40.88 | 85.99 | 10.39 | 11.5 | 12.56 | 71.63 | 41.56 | |
| Llama-3.1-8B-InstructRevision Stage=Round 2, Revision Method=NRLB2026.04 | 39.7 | 85.59 | 14.6 | 11.4 | 12.62 | 62.68 | 34.55 | |
| Qwen3-8BRevision Stage=Round 1, Revision Method=NRLB2026.04 | 37.08 | 85.03 | 13.08 | 12.68 | 14.22 | 66.68 | 36.02 | |
| Qwen3-8BRevision Stage=Round 2, Revision Method=NRLB2026.04 | 34.62 | 84.62 | 11.16 | 12.24 | 13.09 | 68.3 | 38.5 |