Long-context answering with citations on GovReport
82.8Citation RecallGLM-4
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GLM-4Strategy=LAC-S2024.09 | 82.8 | 93.4 | 87.1 | |
| Claude-3-sonnetStrategy=LAC-S2024.09 | 77.4 | 93.9 | 84.1 | |
| LongCite-8BStrategy=LAC-S2024.09 | 74 | 86.6 | 78.5 | |
| GPT-4oStrategy=LAC-S2024.09 | 73.4 | 90.4 | 79.8 | |
| Mistral-Large-InstructStrategy=LAC-S2024.09 | 67.9 | 79.6 | 72.5 | |
| LongCite-9BStrategy=LAC-S2024.09 | 63.4 | 76.5 | 68.2 | |
| Llama-3.1-70B-InstructStrategy=LAC-S2024.09 | 53.4 | 77.5 | 60.7 | |
| Llama-3.1-8B-InstructStrategy=LAC-S2024.09 | 16.2 | 25.3 | 16.8 | |
| GLM-4-9B-chatStrategy=LAC-S2024.09 | 5.7 | 8.2 | 6.3 |