Probability Forecasting on Polymarket 30% held-out evaluation cliques
0.1409Brier Scorehierarchical JCD
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| hierarchical JCDModel=anthropic claude-haiku-4-5-202510012026.05 | 0.1409 | -0.0142 | — | |
| B1Model=anthropic claude-haiku-4-5-202510012026.05 | 0.1423 | — | — | |
| B6Model=anthropic claude-haiku-4-5-202510012026.05 | 0.1435 | — | -0.0115 | |
| B5Model=anthropic claude-haiku-4-5-202510012026.05 | 0.1492 | — | — | |
| B2Model=anthropic claude-haiku-4-5-202510012026.05 | 0.1551 | — | — | |
| hierarchical JCDModel=openai gpt-5.4-mini2026.05 | 0.1796 | -0.0572 | — | |
| B1Model=openai gpt-5.4-mini2026.05 | 0.1798 | — | — | |
| B6Model=openai gpt-5.4-mini2026.05 | 0.1803 | — | -0.0565 | |
| B6Model=openai gpt-5.4-nano2026.05 | 0.1839 | — | -0.0547 | |
| B1Model=meta llama-3.3-70b2026.05 | 0.1872 | — | — | |
| hierarchical JCDModel=meta llama-3.3-70b2026.05 | 0.1872 | -0.0628 | — | |
| B6Model=meta llama-3.3-70b2026.05 | 0.1872 | — | -0.0628 | |
| B1Model=openai gpt-5.4-nano2026.05 | 0.1892 | — | — | |
| hierarchical JCDModel=openai gpt-5.4-nano2026.05 | 0.1892 | -0.0494 | — | |
| B5Model=openai gpt-5.4-mini2026.05 | 0.1898 | — | — | |
| B5Model=openai gpt-5.4-nano2026.05 | 0.1935 | — | — | |
| B5Model=meta llama-3.3-70b2026.05 | 0.197 | — | — | |
| B2Model=openai gpt-5.4-mini2026.05 | 0.2368 | — | — | |
| B2Model=openai gpt-5.4-nano2026.05 | 0.2387 | — | — | |
| B2Model=meta llama-3.3-70b2026.05 | 0.25 | — | — |