Table 3 Prompt sensitivity in GPT-4o and Gemini 1.5 Pro.
From: Multitasking vision language models for vehicle plate recognition with VehiclePaliGemma
Method | Number of correctly | Plate accuracy | |
|---|---|---|---|
Predicted plates | (%) | ||
GPT-4o | Prompt1 | 216 | 83.7 % |
Prompt2 | 222 | 86 % | |
Prompt3 | 227 | 88 % | |
Gemini 1.5 Pro | Prompt1 | 177 | 68.6 % |
Prompt2 | 186 | 72.1 % | |
Prompt3 | 176 | 68.2 % | |
Pre-trained PaliGemma | Prompt2 | 119 | 46.12 % |
Prompt4 | 178 | 69 % |
