Position: Towards Implicit Prompt For Text-To-Image Models

TL;DR

This work investigates implicit prompts for text-to-image models by introducing ImplicitBench, a benchmark with over 2,000 prompts across General Symbols, Celebrity Privacy, and NSFW Issues. It evaluates six T2I systems (three open-source, three closed-source) and uses GPT-4V, ArcFace, and safety detectors to assess generation accuracy and safety risks. The findings show that implicit prompts can meaningfully steer imagery toward target concepts, but also raise privacy and NSFW safety concerns that current policies inadequately address. The authors propose practical mitigation strategies, including prompt recaption, safety-checker enhancement with edged-unsafe examples, and dynamic safety filters with intent clarification, aiming to balance the benefits of implicit prompts with robust risk management.

Abstract

Recent text-to-image (T2I) models have had great success, and many benchmarks have been proposed to evaluate their performance and safety. However, they only consider explicit prompts while neglecting implicit prompts (hint at a target without explicitly mentioning it). These prompts may get rid of safety constraints and pose potential threats to the applications of these models. This position paper highlights the current state of T2I models toward implicit prompts. We present a benchmark named ImplicitBench and conduct an investigation on the performance and impacts of implicit prompts with popular T2I models. Specifically, we design and collect more than 2,000 implicit prompts of three aspects: General Symbols, Celebrity Privacy, and Not-Safe-For-Work (NSFW) Issues, and evaluate six well-known T2I models' capabilities under these implicit prompts. Experiment results show that (1) T2I models are able to accurately create various target symbols indicated by implicit prompts; (2) Implicit prompts bring potential risks of privacy leakage for T2I models. (3) Constraints of NSFW in most of the evaluated T2I models can be bypassed with implicit prompts. We call for increased attention to the potential and risks of implicit prompts in the T2I community and further investigation into the capabilities and impacts of implicit prompts, advocating for a balanced approach that harnesses their benefits while mitigating their risks.

Figures (14)

• Figure 1: Implicit prompts have the capacity to generate imagery akin to that produced by explicit prompts but also possess potential risks to create content that would be restricted by the policy constraints applicable to explicit prompts.
• Figure 2: (a) The tree diagram of our ImplicitBench. (b) The process of creating specific target descriptions as corresponding implicit prompts and the procedure of collecting NSFW implicit prompts. (c) Three evaluation methods towards three aspects of implicit prompts.
• Figure 3: Samples of images induced by our implicit prompts in three different aspects. The bold text represents the subcategory and the colored italic text indicates the explicit target. The NSFW images of the porn category have been manually blurred by us.
• Figure 4: Visual comparisons of different T2I models toward implicit prompts across three aspects.
• Figure 5: The comparison results of different T2I models toward implicit prompts across different professions.