#01
Preds 2.0
- FPR
- 0,42%
- FNR
- 3,1%
- F1-score
- 98,12%
- Accuracy
- 98,68%
Top AI Detectors for Hungarian Text
The benchmark leaderboard was created to provide transparent evaluation and comparison of AI text detector performance on Hungarian text.
2 000texts evaluated.
| # | Model | FPR | FNR | F1-score | Accuracy |
|---|---|---|---|---|---|
| 01 | Preds 2.0 | 0,42% | 3,1% | 98,12% | 98,68% |
| 02 | Hungarian Benchmark v3 | 0,93% | 4,4% | 97,32% | 97,84% |
| 03 | Nyelvi detektor baseline | 1,4% | 5,8% | 96,28% | 96,92% |
#02
Hungarian Benchmark v3
- FPR
- 0,93%
- FNR
- 4,4%
- F1-score
- 97,32%
- Accuracy
- 97,84%
#03
Nyelvi detektor baseline
- FPR
- 1,4%
- FNR
- 5,8%
- F1-score
- 96,28%
- Accuracy
- 96,92%
What do the metrics mean?
These four metrics together measure the reliability of a detector. Lower error rates along with higher F1-scores and accuracy indicate superior performance.
- FPR
- False Positive Rate: indicates how many out of 100 genuine human texts are erroneously flagged as AI. Lower is better.
- FNR
- False Negative Rate: indicates how many out of 100 actual AI-generated texts are missed and classified as human. Lower is better.
- F1-score
- A single score balancing the detection of AI content and the prevention of false alarms. Higher values reflect better overall performance.
- Accuracy
- The overall percentage of evaluated texts correctly categorized by the model. Higher is better.