Economic Strategy Selection on Controlled Economic Sandbox Goods-exchange
31,514Terminal EquityMetaPS-4B V1
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| MetaPS-4B V1Category=LLM and MetaPS Routers, Model size=4B, Version=V12026.06 | 31,514 | 264.14 | 59 | 186.22 | 44.21 | 0 | |
| MetaPS-4B V3Category=LLM and MetaPS Routers, Model size=4B, Version=V32026.06 | 30,454 | 251.89 | 59 | 198.76 | 41.55 | 0.09 | |
| MetaPS-4B V2Category=LLM and MetaPS Routers, Model size=4B, Version=V22026.06 | 27,631 | 219.26 | 57 | 230.85 | 18.69 | 1.07 | |
| Strategy ClassifierCategory=Non-LLM Learned Selectors2026.06 | 26,647 | 207.89 | 53 | 268.37 | 35.14 | 0 | |
| Reward-Ranking SelectorCategory=Non-LLM Learned Selectors2026.06 | 26,002 | 200.45 | 53 | 251.28 | 27.11 | 0 | |
| Qwen-4B BaseCategory=LLM and MetaPS Routers, Model size=4B2026.06 | 23,879 | 175.92 | 41 | 399.45 | 18.25 | 1.21 | |
| Sequence SelectorCategory=Non-LLM Learned Selectors2026.06 | 23,872 | 175.84 | 50 | 339.97 | 18.52 | 1.21 | |
| GPT-5.4Category=LLM and MetaPS Routers2026.06 | 23,554 | 172.16 | 53 | 251.29 | 17.42 | 1.52 | |
| GPT-5.4-miniCategory=LLM and MetaPS Routers2026.06 | 23,249 | 168.64 | 54 | 236.77 | 25.08 | 0.01 | |
| Best Fixed StrategyCategory=Strategy-only Baselines2026.06 | 18,609 | 115.02 | 38 | 443.74 | 12.99 | 3.74 | |
| RL Q-Policy SelectorCategory=Non-LLM Learned Selectors2026.06 | 15,030 | 73.67 | 41 | 680.2 | 8.22 | 5.17 | |
| Random StrategyCategory=Strategy-only Baselines2026.06 | 13,401 | 54.84 | 16.2 | 735.79 | 6.52 | 6.11 |