FinED-Bench - General Knowledge Errors: leaderboard

Metric: F1 (%; general knowledge errors: illegal time, redundant statements, value format errors, missing numerical and non-numerical attribute values; a predicted error counts as correct when the extracted sentence matches or contains the annotated sentence and the error type is classified correctly; the paper's error-detection prompt over each document, reasoning models with thinking on unless marked no thinking). Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.

Top models

#ModelScore
1GPT-4o52.33
2Qwen 3 14B51.41
3Qwen 3 8B47.6
4Qwen 3 14B (Non-reasoning)33.72
5GPT-4o Mini29.69
6DeepSeek R1 0528 Qwen3 8B29.09
7Qwen 3 8B (Non-reasoning)24.93
8Qwen 2.5 7B Instruct14.67
9GPT-3.5 Turbo14.05

Interactive version: theaggregate.ai/benchmark?slug=fined-bench-general-knowledge-errors · How It Works · Data refreshed daily, snapshot 2026-09-26.