Comparison on Loong Set 4: 200K–250K Tokens
55.8LLM ScoreDisco-RAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Disco-RAGBase Model=Llama-3.3-70B2026.01 | 55.8 | 17 | |
| StructRAGCondition=SOTA Results2026.01 | 55.62 | 25 | |
| Disco-RAGBase Model=Qwen2.5-72B2026.01 | 54.97 | 15 | |
| Disco-RAGBase Model=Llama-3.1-8B2026.01 | 53.92 | 12 | |
| RQ-RAGCondition=SOTA Results2026.01 | 40.36 | 0 | |
| Llama-3.3-70BCondition=Standard RAG2026.01 | 34.49 | 2 | |
| Qwen2.5-72BCondition=Standard RAG2026.01 | 32.31 | 1 | |
| Llama-3.3-70BCondition=Full Context2026.01 | 32.22 | 7 | |
| Llama-3.1-8BCondition=Standard RAG2026.01 | 31.9 | 0 | |
| Qwen2.5-72BCondition=Full Context2026.01 | 28.23 | 6 | |
| GraphRAGCondition=SOTA Results2026.01 | 26.67 | 0 | |
| Llama-3.1-8BCondition=Full Context2026.01 | 25.37 | 6 |