Forecasting Resolution on KalshiBench
83.43AccuracyConfidence-Weighted Vote
Evaluation Results
| Method | Links | |
|---|---|---|
| Confidence-Weighted VoteGroup=Architecture A2026.05 | 83.43 | |
| Majority VoteGroup=Architecture A2026.05 | 82.93 | |
| DeepSeekGroup=Single-LLM Baselines2026.05 | 82.42 | |
| GPT-4oGroup=Single-LLM Baselines2026.05 | 82.34 | |
| LlamaGroup=Single-LLM Baselines2026.05 | 81.24 | |
| Final-Round DeepSeekGroup=Architecture B2026.05 | 76.7 | |
| Final-Round LlamaGroup=Architecture B2026.05 | 76.37 | |
| Deliberative FinalGroup=Architecture B2026.05 | 76.11 | |
| Final-Round GPT-4oGroup=Architecture B2026.05 | 74.35 | |
| Claude Opus 4.5Group=Prior Work (KalshiBench)2026.05 | 69.3 |