FalseCite: leaderboard
Metric: Hallucination rate (%) when the false claim is given with no citation; FalseCite prompts built from false FEVER claims and false SciQ answer statements; a GPT-4.1 judge labels each response hallucinated when it accepts or elaborates the false claim, and responses it marks unsure count as not hallucinated; lower is better. Source: arxiv.org. Saturation forecast: Around January 2027. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o Mini | 23.97 | #588 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=falsecite · How It Works · Data refreshed daily, snapshot 2026-10-11.