Chain-of-reasoning on Loong Set 1: 10K–50K Tokens
70.31LLM ScoreLlama-3.3-70B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Llama-3.3-70BCondition=Full Context2026.01 | 70.31 | 37 | |
| Disco-RAGBase Model=Llama-3.3-70B2026.01 | 68.3 | 38 | |
| Disco-RAGBase Model=Llama-3.1-8B2026.01 | 68 | 34 | |
| StructRAGCondition=SOTA Results2026.01 | 67.84 | 34 | |
| Disco-RAGBase Model=Qwen2.5-72B2026.01 | 67.73 | 35 | |
| Qwen2.5-72BCondition=Full Context2026.01 | 66.51 | 36 | |
| Llama-3.3-70BCondition=Standard RAG2026.01 | 66.48 | 36 | |
| Llama-3.1-8BCondition=Full Context2026.01 | 65.66 | 37 | |
| Qwen2.5-72BCondition=Standard RAG2026.01 | 64.67 | 34 | |
| RQ-RAGCondition=SOTA Results2026.01 | 58.96 | 25 | |
| Llama-3.1-8BCondition=Standard RAG2026.01 | 58.76 | 32 | |
| GraphRAGCondition=SOTA Results2026.01 | 54.29 | 43 |