CommunityFact (Closed-Input) - Politics (English): leaderboard
Metric: Macro-F1 (%) over True and False claims on the temporally held-out Politics English test slice (318 True and 183 False standalone claims built from helpful X Community Notes, labels grounded in the note), closed-input (the model sees only the claim and its timestamp, no web search), zero-shot verdict prompt in the claim language; open models greedy, closed models at default settings with reasoning effort medium in their thinking runs; higher is better. Source: arxiv.org. Saturation forecast: Around March 2027. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | aya-expanse-32B | 71.79 |
| 2 | Qwen 3 32B (Non-reasoning) | 68.63 |
| 3 | Grok 4.3 (Medium) | 67.65 |
| 4 | Gemini 2.5 Flash (Medium) | 67.45 |
| 5 | Gemini 2.5 Flash (Non-reasoning) | 65.66 |
| 6 | aya-expanse-8B | 62.71 |
| 7 | Qwen 3 32B | 62.65 |
| 8 | Qwen 3 14B (Non-reasoning) | 62.07 |
| 9 | Qwen 3 14B | 56.99 |
| 10 | GPT-5 Nano (Medium) | 55.47 |
| 11 | Grok 4.3 (Non-reasoning) | 54.93 |
| 12 | EuroLLM-22B-Instruct-2512 | 54.75 |
| 13 | Ministral-8B-Instruct-2410 | 39.59 |
| 14 | GPT-5 Nano (Minimal) | 39.45 |
Interactive version: theaggregate.ai/benchmark?slug=communityfact-closed-input-politics-english · How It Works · Data refreshed daily, snapshot 2026-10-07.