Little bunny, so soft and brown,
Hopping through the field, up and down.
Your ears so long, your tail so small,
You're the cutest creature of them all.
You nibble on the grass, so green and sweet,
And hop away when you see me.
But I'll always be your friend,
Little bunny, until the very end.
更新后的提示
改进后的回答
您能为我写一首关于此图片的兼具描述性和戏剧性的诗歌，并在其中包含地点吗？
Little bunny on the green,
Ears so long, so keen.
What do you see out there,
In the distance, where you stare?
Is it the ocean blue,
Sparkling in the sun's hue?
Or the waves that crash and roar,
A rhythm that you adore?
Little bunny, tell me true,
What captures your attention so?
Is it the salty breeze,
Or the seagulls' cries that please?
Perhaps it's the vastness of it all,
A world so big and small.
A place where dreams can roam,
And adventures can be sown.
Little bunny, I wonder what you dream,
As you sit on the grass, so serene.
Do you yearn to explore the deep,
Or stay on land, where you can leap?
Whatever it may be, little bunny,
Keep that spark of wonder burning brightly.
For in your dreams and aspirations,
Lies a world waiting for your creation.
[[["易于理解","easyToUnderstand","thumb-up"],["解决了我的问题","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["很难理解","hardToUnderstand","thumb-down"],["信息或示例代码不正确","incorrectInformationOrSampleCode","thumb-down"],["没有我需要的信息/示例","missingTheInformationSamplesINeed","thumb-down"],["翻译问题","translationIssue","thumb-down"],["其他","otherDown","thumb-down"]],["最后更新时间 (UTC)：2025-09-04。"],[],[],null,["# Design multimodal prompts\n\nThe Gemini API in Vertex AI lets you send include multimodal inputs like text,\nimages, and video as part of your prompt to Gemini models.\nThis page provides best practices for designing multimodal prompts and how\nto troubleshoot if your prompts aren't working as expected. For general guidance\non prompt design, see\n[our recommended prompt design strategies](/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies),\nor see the following list of multimodal-specific best practices.\n\nPrompt design fundamentals\n--------------------------\n\n### Use specific instructions\n\nClear and detailed prompts provide the best results. If you have a\nspecific output in mind, include that requirement in the prompt to increase the\nlikelihood that you get the correct output.\n\nEven when a prompt seems clear to the person who wrote it, the prompt might be\nunderspecified. To customize model behavior, consider how your prompt\ncould be interpreted and ensure that the instructions you give are specific\nand clear.\n\nIf you included the image of an airport board below as part of your prompt,\nasking the model to just \"describe this image\" could generate a general\ndescription. If you need the model to parse the time and city from the image,\ninclude that request directly in your prompt.\n\n*** ** * ** ***\n\n### Add examples\n\nThe model can accept multiple inputs to use as examples to understand the\noutput you want. Adding these examples can help the model identify the patterns\nand apply the relationship between the given images and responses to the new\nexample. This is called *\"few-shot\" learning*.\n\nIn the following example, the initial output is a full sentence that also\ncontains the country (Brazil). If the preferred output is only the city, not\nthe country, then adding few-shot examples to your prompt can guide\nthe model to respond in that format.\n\n### Split complex tasks into smaller tasks\n\nFor complex tasks like ones that require both visual understanding and\nreasoning, split the task into smaller, more straightforward steps or\ndirectly ask the model to think step by step in the prompt.\n\nMath problems or other types of word problems can also be broken down into smaller tasks.\n\nThe response from the model is incorrect. Try asking the model to parse the problem before answering the prompt.\n\nTo make the response even more streamlined, you can ask the model to think\nthrough the prompt step by step.\n\n### Specify the output format\n\nWhen a certain format is required, you can instruct the model to use that format\nfor its output.\n\nTroubleshoot a multimodal prompt\n--------------------------------\n\nTroubleshooting a prompt can involve improving the language of your prompt,\ndirecting the focus of your prompt, or tuning parameters. These different\napproaches can lead to improved results, especially after determining the root\ncause of the issue.\n\n### Focus on the relevant part of an image\n\nTo improve the response, point out which parts of the image are most\nrelevant to the prompt.\n\n### Tailor the model's response to input\n\nAsk the model to describe the images before performing the task in the prompt.\n\nYou can explicitly ask the model to refer to what's in the image.\n\nThe same tips work for creative output. Explicitly ask the model to include\naspects from the image; this can improve the output.\n\n### Determine how the prompt failed\n\nA prompt can fail because the model did not understand the image at all, or\nbecause it did not perform the correct reasoning steps afterward. To\ndisambiguate between those reasons, ask the model to describe what's in the\nimage.\n\nIn this example, if the model responds with a snack that seems surprising when\npaired with tea (like popcorn), first determine whether the model recognized\nthat the image contains tea.\n\nAsking the model to explain its reasoning can help narrow down which part of\nthe reasoning broke down, if any.\n\n### Tune the sampling parameters\n\nIn each request, you send not only the multimodal prompt but a set of sampling\nparameters to the model. The model can generate different results for different\nparameter values. Experiment with the different parameters to get the best\nvalues for the task. The most commonly adjusted parameters are the following:\n\n- [Temperature](#temperature)\n- [top-P](#top-p)\n\n#### Temperature\n\nTemperature is used for sampling during response generation, which occurs when\ntop-P and top-K are applied. Temperature controls the degree of randomness in\ntoken selection. Lower temperatures are good for prompts that require a more\ndeterministic and less open-ended or creative response, while higher\ntemperatures can lead to more diverse or creative results. A temperature of\n0 is deterministic, meaning that the highest probability response is most\nlikely to be selected.\n\nFor most use cases, start with a temperature of 0.4. If you need more\ncreative results, increase the temperature. If you observe clear\nhallucinations, reduce the temperature.\n\n#### Top-P\n\nTop-P changes how the model selects tokens for output. Tokens are selected\nfrom the most to least probable until the sum of their probabilities equals\nthe top-P value. For example, if tokens A, B, and C have a probability of 0.6,\n0.3, 0.1 and the top-P value is 0.9, then the model will select either A or B\nas the next token by using temperature and excludes C as a candidate.\n\nSpecify a lower value for less random responses and a higher value for more\nrandom responses. The default value of top-P is 1.0.\n\nWhat's next\n-----------\n\n- Try a quickstart tutorial using [Vertex AI Studio](/vertex-ai/generative-ai/docs/start/quickstarts/quickstart) or the [Vertex AI API](/vertex-ai/generative-ai/docs/start/quickstarts/quickstart-multimodal).\n- To get started using the Gemini API in Vertex AI, see the [Gemini API in Vertex AI quickstart](/vertex-ai/generative-ai/docs/start/quickstarts/quickstart-multimodal)."]]
