Chain-of-reasoning on Loong Set 2: 50K–100K Tokens
58.23LLM ScoreDisco-RAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Disco-RAGBase Model=Llama-3.3-70B2026.01 | 58.23 | 22 | |
| Disco-RAGBase Model=Qwen2.5-72B2026.01 | 57.22 | 20 | |
| Llama-3.3-70BCondition=Standard RAG2026.01 | 56.73 | 18 | |
| StructRAGCondition=SOTA Results2026.01 | 54.7 | 19 | |
| Qwen2.5-72BCondition=Standard RAG2026.01 | 53.28 | 16 | |
| Disco-RAGBase Model=Llama-3.1-8B2026.01 | 53.06 | 16 | |
| Llama-3.1-8BCondition=Standard RAG2026.01 | 50.42 | 15 | |
| Llama-3.3-70BCondition=Full Context2026.01 | 50.08 | 10 | |
| Qwen2.5-72BCondition=Full Context2026.01 | 47.69 | 11 | |
| RQ-RAGCondition=SOTA Results2026.01 | 47.6 | 10 | |
| GraphRAGCondition=SOTA Results2026.01 | 46.25 | 12 | |
| Llama-3.1-8BCondition=Full Context2026.01 | 44.49 | 11 |