POLECAT-FOR-ME: leaderboard
Geopolitical event forecasting (POLECAT-FOR-ME, from the ThinkTank-ME paper by Li et al., 2026): given the history of related POLECAT events as (subject, relation, object) triples, the model predicts the missing object of a later query, over 35 countries and regions with the Middle East emphasized. Micro-averaged accuracy in percent over all test queries; higher is better.
Metric: Accuracy (%; micro-average over all test queries of 35 countries and regions: the predicted object entity of a (subject, relation, ?) event query). Source: arxiv.org. Saturation forecast: Around March 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 23.5 | #741 |
| 2 | GPT-4o Mini | 22.7 | #1627 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=polecat-for-me · How It Works · Data refreshed daily, snapshot 2026-11-12.