Financial Numerical Reasoning on FinQA (test)
84.66Execution AccuracyDSA-7B
Evaluation Results
| Method | Links | |
|---|---|---|
| DSA-7BModel Scale=7B2026.05 | 84.66 | |
| DSA-3BModel Scale=3B2026.05 | 79.12 | |
| MOCA-AGENTBackbone=Qwen3.6-27B, Type=Open-source LLMs2026.06 | 78.29 | |
| FINDERBackbone=Mistral-7B Instruct v0.22026.05 | 75.32 | |
| Fino1-14BType=Open-source LLMs2026.06 | 74.18 | |
| DSA-1.5BModel Scale=1.5B2026.05 | 73.5 | |
| Qwen2.5-72BType=Open-source LLMs2026.06 | 73.38 | |
| DeepSeek-V3Type=Proprietary LLMs2026.06 | 73.2 | |
| Qwen2.5-32BType=Open-source LLMs2026.06 | 73.11 | |
| Fino1-7BType=Open-source LLMs2026.06 | 73.03 | |
| GPT-4oType=Proprietary LLMs2026.06 | 72.49 | |
| APOLLO2026.05 | 71.07 | |
| ENCORE2026.05 | 69.4 | |
| PoT-GPT-4Backbone=GPT-42026.05 | 69.38 | |
| GPT-4.5Type=Proprietary LLMs2026.06 | 68.94 | |
| Llama-3.3-70BType=Open-source LLMs2026.06 | 68.15 | |
| CBR-Ren2026.05 | 67.81 | |
| DS-R1-Distill-Llama-70BType=Open-source LLMs2026.06 | 66.73 | |
| DeepSeek-R1Type=Proprietary LLMs2026.06 | 65.13 | |
| ArgRecog2026.05 | 64.86 | |
| Qwen3-32BType=Open-source LLMs2026.06 | 64.15 | |
| FinQANet2026.05 | 61.24 | |
| GPT-o3-miniType=Proprietary LLMs2026.06 | 60.87 | |
| Qwen2.5-7B-InsZero-shot=true, Model Scale=7B2026.05 | 60.8 | |
| GPT-o1-previewType=Proprietary LLMs2026.06 | 49.07 | |
| Qwen2.5-3B-InsZero-shot=true, Model Scale=3B2026.05 | 47.5 | |
| Qwen2.5-1.5B-InsZero-shot=true, Model Scale=1.5B2026.05 | 34.22 |