Financial Document QA on FinQA (test)
76.81Execution AccuracyFinAgent-RAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FinAgent-RAGBackbone=DeepSeek-V32026.05 | 76.81 | 73.58 | |
| IterRAGCategory=Agentic, Backbone=DeepSeek-V32026.05 | 67.83 | 64.76 | |
| CRAGCategory=Self-correcting, Backbone=DeepSeek-V32026.05 | 66.21 | 63.14 | |
| Self-RAGCategory=Self-correcting, Backbone=DeepSeek-V32026.05 | 64.75 | 61.68 | |
| ReActCategory=Agentic, Backbone=DeepSeek-V32026.05 | 63.18 | 60.05 | |
| Advanced RAGCategory=Standard RAG, Backbone=DeepSeek-V32026.05 | 61.52 | 58.34 | |
| FinMACategory=Non-retrieval2026.05 | 59.63 | 56.71 | |
| Naive RAGCategory=Standard RAG, Backbone=DeepSeek-V32026.05 | 57.84 | 54.62 | |
| Zero-shot LLMCategory=Non-retrieval, Backbone=DeepSeek-V32026.05 | 53.27 | 50.14 |