Financial Reasoning on ConvFinQA
85.7AccuracyGPT-5 mini-high
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GPT-5 mini-highReasoning Capability=with Reasoning, Domain Focus=General2026.03 | 85.7 | 77.1 | |
| Qwen3-32BReasoning Capability=without Reasoning, Domain Focus=General2026.03 | 82.5 | 74.7 | |
| Qwen3-4B-ThinkingReasoning Capability=with Reasoning, Domain Focus=General2026.03 | 80.6 | 72.5 | |
| ODA-Fin-RL-8BDomain Focus=Financial2026.03 | 80.4 | 74.6 | |
| Gemini 2.5 Flash-LiteReasoning Capability=with Reasoning, Domain Focus=General2026.03 | 79.5 | 67.7 | |
| Qwen3-8BReasoning Capability=without Reasoning, Domain Focus=General2026.03 | 78.8 | 71.5 | |
| ODA-Fin-SFT-8BDomain Focus=Financial2026.03 | 78.3 | 72.1 | |
| Fin-R1Domain Focus=Financial2026.03 | 76.2 | 61.4 | |
| Dianjin-R1-7BDomain Focus=Financial2026.03 | 74.6 | 70.3 | |
| Qwen2.5-7B-InstructReasoning Capability=without Reasoning, Domain Focus=General2026.03 | 72.1 | 68 | |
| Qwen2.5-7BReasoning Capability=without Reasoning, Domain Focus=General2026.03 | 69.6 | 47.8 | |
| Llama-3.1-8B-InstructReasoning Capability=without Reasoning, Domain Focus=General2026.03 | 68.2 | 56.9 | |
| ReasonAnyBackbone=Qwen2.5-7B family2026.01 | 55.85 | — | |
| LEDBackbone=Qwen2.5-7B family2026.01 | 51.01 | — | |
| FinanceBackbone=Qwen2.5-7B family2026.01 | 50.35 | — | |
| ReasoningBackbone=Qwen2.5-7B family2026.01 | 36.54 | — | |
| LinearBackbone=Qwen2.5-7B family2026.01 | 34.02 | — | |
| TIESBackbone=Qwen2.5-7B family2026.01 | 18.87 | — | |
| FuseLLMBackbone=Qwen2.5-7B family2026.01 | 18.44 | — | |
| DAREBackbone=Qwen2.5-7B family2026.01 | 17.75 | — | |
| Task ArithmeticBackbone=Qwen2.5-7B family2026.01 | 17.43 | — | |
| Xuanyuan-6B-ChatDomain Focus=Financial2026.03 | 10.9 | 36.7 | |
| Plutus-8B-InstructDomain Focus=Financial2026.03 | 9.3 | 36.2 |