Investment decision alignment on P3 v1 (test)
1.59Overall MSEInvestAgent
Evaluation results
| Method | Links | ||
|---|---|---|---|
| InvestAgentBackbone=Llama-3.1, Training Stage=Post-InvestAlign2025.07 | 1.59 | -6,103 | |
| InvestAgentBackbone=GPT-3.5, Training Stage=Post-InvestAlign2025.07 | 1.72 | -6,126 | |
| InvestAgentBackbone=Qwen-2, Training Stage=Post-InvestAlign2025.07 | 2.16 | -4,559 | |
| Qwen-2Training Stage=Pre-SFT2025.07 | 3.97 | — | |
| Llama-3.1Training Stage=Pre-SFT2025.07 | 4.08 | — | |
| GPT-3.5Training Stage=Pre-SFT2025.07 | 4.44 | — |