Long-context Understanding on LongBench NarrativeQA
27.4F1 ScoreBase+ctx
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Base+ctxLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 27.4 | 26.56 | |
| Base+ctxLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=Zero-shot2026.06 | 22.67 | 22.37 | |
| Doc2AtomLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=Zero-shot2026.06 | 20.58 | 20.27 | |
| Doc2AtomLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 17.87 | 17.82 | |
| D2LckptLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=Zero-shot2026.06 | 13.31 | 13.27 | |
| D2LckptLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 11.67 | 11.2 | |
| D2LatomLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=Zero-shot2026.06 | 6.21 | 6.15 | |
| D2LrawLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=Zero-shot2026.06 | 4.64 | 4.64 | |
| D2LatomLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 3.46 | 3.46 | |
| D2LrawLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 2.83 | 2.83 |