Chain-of-reasoning on Loong Set 3: 100K–200K Tokens
0.5217LLM ScoreDisco-RAG
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Disco-RAGBase Model=Llama-3.3-70B2026.01 | 0.5217 | 0.15 | |
| Disco-RAGBase Model=Qwen2.5-72B2026.01 | 0.512 | 0.13 | |
| Disco-RAGBase Model=Llama-3.1-8B2026.01 | 0.5031 | 0.08 | |
| Llama-3.3-70BCondition=Standard RAG2026.01 | 0.501 | 0.05 | |
| Llama-3.3-70BCondition=Full Context2026.01 | 0.4978 | 0.01 | |
| StructRAGCondition=SOTA Results2026.01 | 0.4973 | 0.13 | |
| Qwen2.5-72BCondition=Standard RAG2026.01 | 0.4909 | 0.02 | |
| Llama-3.1-8BCondition=Standard RAG2026.01 | 0.4905 | 0 | |
| Qwen2.5-72BCondition=Full Context2026.01 | 0.4847 | 0.01 | |
| GraphRAGCondition=SOTA Results2026.01 | 0.4333 | 0.17 | |
| RQ-RAGCondition=SOTA Results2026.01 | 0.3679 | 0.07 | |
| Llama-3.1-8BCondition=Full Context2026.01 | 0.3562 | 0 |