Multi-round Investment Game Evaluation (ALIGN agents protocol, 5 seeds)
54.54Discounted ReturnDeepSeek-V3.1 Reasoner
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DeepSeek-V3.1 ReasonerAgent Category=Reasoning Models, Protocol=ALIGN2026.02 | 54.54 | 13.93 | 70 | 47 | 0.33 | |
| DeepSeek-V3.1 ChatAgent Category=Chat Models, Protocol=ALIGN2026.02 | 32.74 | 8.36 | 42 | 45 | 0.29 | |
| GPT-4o MiniAgent Category=Chat Models, Protocol=ALIGN2026.02 | 17.18 | 4.38 | 47 | 77 | 0.22 | |
| o4-miniAgent Category=Reasoning Models, Protocol=ALIGN2026.02 | 11.05 | 2.81 | 32 | 25 | 0.57 | |
| LLaMA 4 MaverickAgent Category=Chat Models, Protocol=ALIGN2026.02 | 9.27 | 2.36 | 24 | 30 | 0.19 | |
| Qwen3-235B-InstructAgent Category=Reasoning Models, Protocol=ALIGN2026.02 | 4.2 | 1.07 | 15 | 6 | 0.12 | |
| Gemini 2.5 Flash-LiteAgent Category=Chat Models, Protocol=ALIGN2026.02 | 0 | 0 | 0 | 0 | 0 | |
| Kimi-K2-InstructAgent Category=Reasoning Models, Protocol=ALIGN2026.02 | 0 | 0 | 0 | 0 | 0 |