Decision-making on 30 synthetic decision-making rounds (evaluation)
2.55Mean RankLark Full
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Lark FullAvg. Cost ($/task)=0.0162025.10 | 2.55 | 2.17 | 29.4 | 26.34 | |
| GPT o3Avg. Cost ($/task)=0.0162025.10 | 4.3 | 3.1 | 28.8 | 26.3 | |
| Qwen3 Next 80B A3B ThinkingAvg. Cost ($/task)=0.0062025.10 | 4.4 | 2.98 | 27.9 | 23.99 | |
| Lark NoPenaltyAvg. Cost ($/task)=0.0112025.10 | 6.15 | 5.41 | 27.2 | 24.28 | |
| GPT o4-miniAvg. Cost ($/task)=0.0092025.10 | 6.5 | 5.44 | 27.2 | 24.64 | |
| Lark NoRankedChoiceVotingAvg. Cost ($/task)=0.0062025.10 | 6.55 | 5.3 | 27 | 21.69 | |
| GPT 4.1Avg. Cost ($/task)=0.0142025.10 | 7.25 | 5.61 | 26.9 | 23.95 | |
| Lark NoPlasticityAvg. Cost ($/task)=0.0102025.10 | 8.3 | 7.24 | 26.2 | 22.7 | |
| GPT5 MiniAvg. Cost ($/task)=0.0072025.10 | 8.6 | 5.67 | 25.4 | 20 | |
| Lark NoMutationNoDuplicationAvg. Cost ($/task)=0.0062025.10 | 8.95 | 7.9 | 25.9 | 22.5 | |
| GPT 4oAvg. Cost ($/task)=0.0132025.10 | 9.5 | 7.17 | 25.4 | 18.6 | |
| GPT o3-miniAvg. Cost ($/task)=0.0102025.10 | 9.6 | 8.4 | 25.6 | 22.67 | |
| Deepseek V3.1 (aka Lark Base)Avg. Cost ($/task)=0.0022025.10 | 10.6 | 8.61 | 23.7 | 15.6 | |
| GPT5 NanoAvg. Cost ($/task)=0.0022025.10 | 11.75 | 10.01 | 23 | 18 |