Long-context Understanding on LongBench (QASPER)
45.14F1 ScoreBase+ctx
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Base+ctxLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 45.14 | 42.93 | |
| Base+ctxLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=Zero-shot2026.06 | 37.05 | 35.58 | |
| Doc2AtomLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 31.8 | 31.46 | |
| Doc2AtomLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=Zero-shot2026.06 | 29.85 | 29.26 | |
| D2LckptLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 21.71 | 20.35 | |
| D2LckptLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=Zero-shot2026.06 | 19.81 | 18.87 | |
| D2LatomLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=Zero-shot2026.06 | 17.1 | 16.68 | |
| D2LrawLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 14.84 | 14.64 | |
| D2LatomLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 14.56 | 14.12 | |
| D2LrawLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=Zero-shot2026.06 | 12.29 | 11.89 |