Financial Reasoning on FinAgent
66Accuracy (Steps)Supervisor-Committee
Evaluation Results
| Method | Links | |
|---|---|---|
| Supervisor-CommitteeBackbone=Gemini 3.1 Pro, Evaluation Protocol=LLM-as-a-Judge2026.05 | 66 | |
| Supervisor-Committee + DoCBackbone=Gemini 3.1 Pro, Evaluation Protocol=LLM-as-a-Judge2026.05 | 58.7 | |
| Supervisor + AgentsBackbone=Gemini 3.1 Pro, Evaluation Protocol=LLM-as-a-Judge2026.05 | 52 | |
| LLM + Tools (ReAct)Backbone=Gemini 3.1 Pro, Evaluation Protocol=LLM-as-a-Judge2026.05 | 46 | |
| LLMBackbone=Gemini 3.1 Pro, Evaluation Protocol=LLM-as-a-Judge2026.05 | 24 |