Preprint
Aug 2026
A Benchmark Dataset for MLLM-Generated Image Detection: GPT Image2&Nano Banana2
A benchmark dataset for detecting images generated by MLLMs is constructed and a structural-artifact-prior-guided dual-stream prompt framework (SAP-DSP) is proposed as a strong baseline, using dual-stream prompt learning and structure-aware routing fusion to improve representation learning.
Zi-Rui Zhang, Yin-Bo Yu, Donghai Guan et al.
· 1 citation
· ⚡1