Financial Question Answering on ConvFinQA (test)
78.46Execution AccuracyFinAgent-RAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FinAgent-RAGBackbone=DeepSeek-V32026.05 | 78.46 | 81.93 | |
| IterRAGBackbone=DeepSeek-V32026.05 | 69.14 | 72.67 | |
| CRAGBackbone=DeepSeek-V32026.05 | 66.72 | 70.28 | |
| Self-RAGBackbone=DeepSeek-V32026.05 | 64.83 | 68.14 | |
| ReActBackbone=DeepSeek-V32026.05 | 62.47 | 65.93 | |
| Advanced RAGBackbone=DeepSeek-V32026.05 | 60.34 | 63.87 | |
| FinMA2026.05 | 57.92 | 61.37 | |
| Naive RAGBackbone=DeepSeek-V32026.05 | 56.14 | 59.82 | |
| Zero-shot LLMBackbone=DeepSeek-V32026.05 | 51.78 | 55.43 |