Enables verification of claimed text-to-image models through boundary-aware prompts that trigger model-specific instability.
March 30, 2026
Original Paper
Verify Claimed Text-to-Image Models via Boundary-Aware Prompt Optimization
arXiv · 2603.26328
The Takeaway
As third-party API platforms proliferate, verifying if an API is actually using 'Stable Diffusion' or 'DALL-E' is difficult; this method identifies prompts near semantic boundaries that serve as unique fingerprints. It provides a robust, reference-free tool for model owners to protect IP and for users to ensure service quality.
From the abstract
As Text-to-Image (T2I) generation becomes widespread, third-party platforms increasingly integrate multiple model APIs for convenient image creation. However, false claims of using official models can mislead users and harm model owners' reputations, making model verification essential to confirm whether an API's underlying model matches its claim. Existing methods address this by using verification prompts generated by official model owners, but the generation relies on multiple reference model