Long-context Understanding on LongBench MFQA-en
47.18F1 ScoreBase+ctx
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Base+ctxLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 47.18 | 45.6 | |
| Base+ctxLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=Zero-shot2026.06 | 39.41 | 38.31 | |
| Doc2AtomLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 34.29 | 33.31 | |
| D2LckptLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 26.69 | 24.31 | |
| Doc2AtomLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=Zero-shot2026.06 | 26 | 25.83 | |
| D2LckptLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=Zero-shot2026.06 | 20.92 | 19.7 | |
| D2LrawLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 10.7 | 9.92 | |
| D2LatomLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 9.73 | 8.94 | |
| D2LrawLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=Zero-shot2026.06 | 7.14 | 6.91 | |
| D2LatomLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=Zero-shot2026.06 | 4.68 | 4.68 |