Financial Numerical Reasoning on FinQA (dev)
84.71Execution AccuracyDSA-7B
Evaluation results
| Method | Links | |
|---|---|---|
| DSA-7BModel Scale=7B2026.05 | 84.71 | |
| DSA-3BModel Scale=3B2026.05 | 79.45 | |
| FINDERBackbone=Mistral-7B Instruct v0.22026.05 | 77.13 | |
| DSA-1.5BModel Scale=1.5B2026.05 | 73.05 | |
| APOLLO2026.05 | 72.91 | |
| ENCORE2026.05 | 71.6 | |
| PoT-GPT-4Backbone=GPT-42026.05 | 71.05 | |
| CBR-Ren2026.05 | 68.4 | |
| ArgRecog2026.05 | 67.5 | |
| Qwen2.5-7B-InsZero-shot=true, Model Scale=7B2026.05 | 61.5 | |
| FinQANet2026.05 | 61.22 | |
| Qwen2.5-3B-InsZero-shot=true, Model Scale=3B2026.05 | 48.3 | |
| Qwen2.5-1.5B-InsZero-shot=true, Model Scale=1.5B2026.05 | 35.1 |