Long-document lay summarization on SciNews
21.11RL ScoreDisco-RAG
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Disco-RAGBackbone=Llama-3.3-70B, Evaluation Setup=Disco-RAG2026.01 | 21.11 | 65.67 | 44.37 | 69.48 | |
| RSTformerEvaluation Setup=SOTA Results2026.01 | 20.12 | 62.8 | 41.56 | — | |
| Disco-RAGBackbone=Qwen2.5-72B, Evaluation Setup=Disco-RAG2026.01 | 20.1 | 64.83 | 41.48 | 66.3 | |
| SingleTurnPlanEvaluation Setup=SOTA Results2026.01 | 19.68 | — | — | — | |
| Disco-RAGBackbone=Llama-3.1-8B, Evaluation Setup=Disco-RAG2026.01 | 19.25 | 63.47 | 40.25 | 63.35 | |
| Llama-3.3-70BBackbone=Llama-3.3-70B, Evaluation Setup=Standard RAG2026.01 | 18.17 | 61.37 | 37.74 | 60.39 | |
| Qwen2.5-72BBackbone=Qwen2.5-72B, Evaluation Setup=Standard RAG2026.01 | 18.09 | 61.28 | 38.32 | 60.12 | |
| Llama-3.3-70BBackbone=Llama-3.3-70B, Evaluation Setup=Full context2026.01 | 17.19 | 61.03 | 37.65 | 54.73 | |
| Llama-3.1-8BBackbone=Llama-3.1-8B, Evaluation Setup=Standard RAG2026.01 | 17.12 | 60.35 | 38.01 | 55.26 | |
| Qwen2.5-72BBackbone=Qwen2.5-72B, Evaluation Setup=Full context2026.01 | 17 | 60.41 | 37.62 | 55.03 | |
| Llama-3.1-8BBackbone=Llama-3.1-8B, Evaluation Setup=Full context2026.01 | 15.33 | 59.27 | 35.43 | 48.31 | |
| Plan-InputEvaluation Setup=SOTA Results2026.01 | — | 65.32 | — | 72.4 |