Evaluation on Goods-exchange Sandbox
31,514Terminal EquityMetaPS-4B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MetaPS-4BParameters=4B2026.06 | 31,514 | 264.14 | 59 | |
| MetaPS-9BParameters=9B2026.06 | 30,455 | 251.9 | 56 | |
| Strategy ClassifierType=Non-LLM Selector2026.06 | 26,647 | 207.89 | 53 | |
| Reward-Ranking SelectorType=Non-LLM Selector2026.06 | 26,002 | 200.45 | 53 | |
| Qwen3.5-4BBackbone=Qwen3.5, Parameters=4B2026.06 | 23,879 | 175.92 | 41 | |
| Sequence SelectorType=Non-LLM Selector2026.06 | 23,872 | 175.84 | 50 | |
| GPT-5.4Model=GPT-5.42026.06 | 23,554 | 172.16 | 53 | |
| GPT-5.4-miniModel=GPT-5.4, Variant=mini2026.06 | 23,249 | 168.64 | 54 | |
| Qwen3.5-9BBackbone=Qwen3.5, Parameters=9B2026.06 | 18,912 | 118.53 | 23 | |
| Best Fixed StrategyStrategy type=Baseline2026.06 | 18,609 | 115.02 | 38 | |
| RL Q-Policy SelectorType=Non-LLM Selector2026.06 | 15,030 | 73.67 | 41 | |
| Random StrategyStrategy type=Baseline2026.06 | 13,401 | 54.84 | 16.2 |