DiffMark enables multi-bit watermarking that is transferable across different frozen diffusion models with a 45x speedup over current methods.
March 24, 2026
Original Paper
Transferable Multi-Bit Watermarking Across Frozen Diffusion Models via Latent Consistency Bridges
arXiv · 2603.20304
The Takeaway
By using Latent Consistency Models as a 'bridge' for backpropagation, it bypasses the need for costly DDIM inversion. This allows for fast, single-pass watermark detection (16ms) that doesn't break when switching model versions or architectures.
From the abstract
As diffusion models (DMs) enable photorealistic image generation at unprecedented scale, watermarking techniques have become essential for provenance establishment and accountability. Existing methods face challenges: sampling-based approaches operate on frozen models but require costly $N$-step Denoising Diffusion Implicit Models (DDIM) inversion (typically N=50) for zero-bit-only detection; fine-tuning-based methods achieve fast multi-bit extraction but couple the watermark to a specific model