I’ve been trying to wrap my head around solving this one. First of all are think/reasoning tags model-specific (baked-in) or is that a setting you can change when hosting your own models? What if I don’t want to pass reasoning content to my users which I have noticed is usually between < think > t…

The <think> tag is a model-specific “convention,” not a strict or special requirement. Therefore, while there is some risk, other inputs and outputs carry equivalent risks. It would be safer to always inspect and process inputs passed to the API and outputs received from the API. Short direct ans…

Prompt Injection concern with think tags

Models

jydarn December 5, 2025, 8:19pm 3

Thanks for the detailed response! That gave me the perspective I needed.

1 Like

Topic		Replies	Views
Inference API detailed request Beginners	5	2379	September 11, 2020
Labels in language modeling: which tokens to set to -100? Beginners	1	3515	November 30, 2020
Masked language modelling with specific entities or POS 🤗Transformers	0	213	July 21, 2021
Feature request: model tags Site Feedback	2	613	August 20, 2020
Inference Hyperparameters Amazon SageMaker	29	4903	October 8, 2021

Prompt Injection concern with think tags

Related topics