Search papers, labs, and topics across Lattice.
University of Electronic Science and Technology of China
4
0
5
Dodging generator-specific fingerprints is now possible: FGINet leverages frequency-aware gating to fuse semantic and artifact cues, significantly boosting generalization in AI-generated image detection.
Existing synthetic image detectors fail to generalize to new generators, but HiMix closes the gap by learning artifact-aware representations across diverse distributions.
Video-LLMs hallucinate because they fixate on a single "anchor frame," but a simple decoder-side attention fix can dramatically improve grounding without retraining.
Achieve >95% forget quality in LLMs with minimal side effects by isolating and unlearning tokens within target subdomains using asymmetric LoRA.
