Today I'm sharing a new tool I've built: the Dumpling Counter Prompt.
This prompt is designed for image-based counting tasks. Its core function is to provide a structured approach for large language models (LLMs) to process visual information and deliver an accurate count of specific items within an image.
The prompt guides the model to reason through the counting process step-by-step. This encourages the model to break down the task, which can improve the reliability of the final count. A key instruction within the prompt is to include notes on any ambiguities. If the model encounters elements that are difficult to categorize or count precisely, it's directed to flag these instances. This helps users understand the confidence level of the count and where potential discrepancies might arise.
This prompt is intended for developers or users who need to automate counting from images and require not just a number, but also insight into the counting process and any edge cases.
Top comments (0)