Long-Document Question Answering on LongBench
30.4NarQA ScoreQwen-LiteCoST
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen-LiteCoSTBase model=Qwen2-7B, Parameters=7B, Framework=LiteCoST2026.03 | 30.4 | 44.64 | 68.39 | 65.73 | |
| GPT-4o2026.03 | 28.68 | 43.39 | 67.68 | 68.29 | |
| LLaMA-LiteCoSTBase model=LLaMA-3.2-3B, Parameters=3B, Framework=LiteCoST2026.03 | 27.24 | 41.37 | 66.86 | 67.52 | |
| GPT-4o-mini2026.03 | 24.38 | 40.28 | 65.03 | 65.15 | |
| Qwen2-7BParameters=7B2026.03 | 19.49 | 35.67 | 45.06 | 41.4 | |
| LLaMA-3.2-3BParameters=3B2026.03 | 16.94 | 34.46 | 54.92 | 51.82 |