List your company

Long-document lay summarization on SciNews

21.11RL ScoreDisco-RAG
15.098816.659418.2219.7806Jan 7, 2026

Evaluation results

MethodLinks
Disco-RAGBackbone=Llama-3.3-70B, Evaluation Setup=Disco-RAG2026.0121.1165.6744.3769.48
RSTformerEvaluation Setup=SOTA Results2026.0120.1262.841.56—
Disco-RAGBackbone=Qwen2.5-72B, Evaluation Setup=Disco-RAG2026.0120.164.8341.4866.3
SingleTurnPlanEvaluation Setup=SOTA Results2026.0119.68———
Disco-RAGBackbone=Llama-3.1-8B, Evaluation Setup=Disco-RAG2026.0119.2563.4740.2563.35
Llama-3.3-70BBackbone=Llama-3.3-70B, Evaluation Setup=Standard RAG2026.0118.1761.3737.7460.39
Qwen2.5-72BBackbone=Qwen2.5-72B, Evaluation Setup=Standard RAG2026.0118.0961.2838.3260.12
Llama-3.3-70BBackbone=Llama-3.3-70B, Evaluation Setup=Full context2026.0117.1961.0337.6554.73
Llama-3.1-8BBackbone=Llama-3.1-8B, Evaluation Setup=Standard RAG2026.0117.1260.3538.0155.26
Qwen2.5-72BBackbone=Qwen2.5-72B, Evaluation Setup=Full context2026.011760.4137.6255.03
Llama-3.1-8BBackbone=Llama-3.1-8B, Evaluation Setup=Full context2026.0115.3359.2735.4348.31
Plan-InputEvaluation Setup=SOTA Results2026.01—65.32—72.4