Financial Question Answering on TAT-QA (test)
74.96Execution AccuracyFinAgent-RAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FinAgent-RAGBackbone=DeepSeek-V32026.05 | 74.96 | 78.13 | |
| IterRAGBackbone=DeepSeek-V32026.05 | 69.34 | 72.51 | |
| CRAGBackbone=DeepSeek-V32026.05 | 67.58 | 70.74 | |
| Self-RAGBackbone=DeepSeek-V32026.05 | 66.12 | 69.23 | |
| ReActBackbone=DeepSeek-V32026.05 | 64.35 | 67.48 | |
| Advanced RAGBackbone=DeepSeek-V32026.05 | 62.67 | 65.89 | |
| Naive RAGBackbone=DeepSeek-V32026.05 | 58.93 | 62.14 | |
| FinMA2026.05 | 57.41 | 60.73 | |
| Zero-shot LLMBackbone=DeepSeek-V32026.05 | 54.18 | 57.32 |