Chain-of-reasoning on Loong Set 4: 200K–250K Tokens
36.17LLM ScoreDisco-RAG
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Disco-RAGBase Model=Qwen2.5-72B2026.01 | 36.17 | 0.06 | |
| Disco-RAGBase Model=Llama-3.3-70B2026.01 | 36.06 | 0.06 | |
| Disco-RAGBase Model=Llama-3.1-8B2026.01 | 36 | 0.03 | |
| StructRAGCondition=SOTA Results2026.01 | 35.71 | 0.05 | |
| RQ-RAGCondition=SOTA Results2026.01 | 34.69 | 0 | |
| GraphRAGCondition=SOTA Results2026.01 | 33.67 | 0.33 | |
| Llama-3.3-70BCondition=Standard RAG2026.01 | 31.33 | 0.02 | |
| Llama-3.3-70BCondition=Full Context2026.01 | 30.17 | 0 | |
| Qwen2.5-72BCondition=Standard RAG2026.01 | 30.02 | 0.01 | |
| Llama-3.1-8BCondition=Standard RAG2026.01 | 29.92 | 0 | |
| Qwen2.5-72BCondition=Full Context2026.01 | 28.48 | 0 | |
| Llama-3.1-8BCondition=Full Context2026.01 | 26.76 | 0 |