Question Answering on LongBench
73.5F1 ScoreTri-RAG
Evaluation results
| Method | Links | |
|---|---|---|
| Tri-RAGGenerator Backbone=Qwen-Long2026.04 | 73.5 | |
| Tri-RAGGenerator Backbone=GPT-4o-mini2026.04 | 70.9 | |
| AflowGenerator Backbone=Qwen-Long2026.04 | 70.8 | |
| Self-ActGenerator Backbone=GPT-4o-mini2026.04 | 68.9 | |
| Self-ActGenerator Backbone=Qwen-Long2026.04 | 68.3 | |
| Tri-RAGGenerator Backbone=GPT-4.1-nano2026.04 | 68.2 | |
| AflowGenerator Backbone=GPT-4.1-nano2026.04 | 66.1 | |
| Self-ActGenerator Backbone=GPT-4.1-nano2026.04 | 64.5 | |
| ReActGenerator Backbone=Qwen-Long2026.04 | 64.2 | |
| Tri-RAGGenerator Backbone=Gemma-2-27B2026.04 | 63.3 | |
| Tri-RAGGenerator Backbone=Qwen-Max2026.04 | 62.8 | |
| ReActGenerator Backbone=GPT-4o-mini2026.04 | 62.7 | |
| AflowGenerator Backbone=Qwen-Max2026.04 | 61.2 | |
| AflowGenerator Backbone=GPT-4o-mini2026.04 | 61 | |
| AflowGenerator Backbone=Gemma-2-27B2026.04 | 60.8 | |
| ReActGenerator Backbone=GPT-4.1-nano2026.04 | 59.7 | |
| DirectGenerator Backbone=Qwen-Long2026.04 | 59.3 | |
| Self-ActGenerator Backbone=Qwen-Max2026.04 | 58.7 | |
| DirectGenerator Backbone=GPT-4o-mini2026.04 | 58.1 | |
| Self-ActGenerator Backbone=Gemma-2-27B2026.04 | 56.4 | |
| DirectGenerator Backbone=GPT-4.1-nano2026.04 | 56.3 | |
| ReActGenerator Backbone=Gemma-2-27B2026.04 | 52.5 | |
| ReActGenerator Backbone=Qwen-Max2026.04 | 52.3 | |
| DirectGenerator Backbone=Gemma-2-27B2026.04 | 49.3 | |
| DirectGenerator Backbone=Qwen-Max2026.04 | 47.2 |