FinED-Bench - Financial Domain Knowledge Errors: leaderboard
FinED-Bench (2026) tests whether LLMs can find and classify errors injected into real 2025 Chinese financial documents from nine scenarios. This board scores the domain knowledge errors (terminology misuse, incorrect legal references, ambiguous expressions, unit errors and omitted financial elements). F1 in percent; higher is better.
Metric: F1 (%; financial domain knowledge errors: terminology misuse, incorrect legal reference, ambiguous expression, numerical unit error, omitted financial element; 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
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 47.36 | #741 |
| 2 | Qwen 3 14B | 41.1 | #1235 |
| 3 | Qwen 3 8B | 38.6 | #1838 |
| 4 | Qwen 3 14B (Non-reasoning) | 26.63 | #1235 (Qwen 3 14B) |
| 5 | DeepSeek R1 0528 Qwen3 8B | 22.82 | #2101 |
| 6 | Qwen 3 8B (Non-reasoning) | 18.06 | #1838 (Qwen 3 8B) |
| 7 | Qwen 2.5 7B Instruct | 8.33 | #2583 |
| 8 | GPT-4o Mini | 7.34 | #1627 |
| 9 | GPT-3.5 Turbo | 3.65 | #2773 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=fined-bench-financial-domain-knowledge-errors · How It Works · Data refreshed daily, snapshot 2026-11-12.