Long-document and Multi-hop Question Answering on 2WikiMultiHopQA
50.39F1 ScoreD2Latom
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| D2LatomLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=In-domain2026.06 | 50.39 | 50.25 | |
| D2LrawLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=In-domain2026.06 | 49.84 | 49.79 | |
| D2LatomLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=In-domain2026.06 | 47.43 | 47.43 | |
| D2LrawLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=In-domain2026.06 | 46.06 | 46.03 | |
| Base+ctxLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=In-domain2026.06 | 45.31 | 45.27 | |
| Doc2AtomLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=In-domain2026.06 | 39.62 | 39.62 | |
| Base+ctxLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=In-domain2026.06 | 37.69 | 37.69 | |
| D2LckptLLM Backbone=Gemma-2-2B-It, Evaluation Protocol=In-domain2026.06 | 36.08 | 36.08 | |
| Doc2AtomLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=In-domain2026.06 | 31.74 | 31.74 | |
| D2LckptLLM Backbone=Qwen3-4B-Instruct, Evaluation Protocol=In-domain2026.06 | 11.25 | 11.17 |