Sitemap
A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.
Pages
Posts
publications
UIBDiffusion: Universal Imperceptible Backdoor Attack for Diffusion Models
Published in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025
A universal, imperceptible backdoor trigger for diffusion models based on universal adversarial perturbations, achieving high attack success rate while preserving generation quality and evading state-of-the-art backdoor defenses (Elijah, TERD).
Recommended citation: Han, Y., Zhao, B., Chu, R., Luo, F., Sikdar, B., & Lao, Y. (2025). "UIBDiffusion: Universal Imperceptible Backdoor Attack for Diffusion Models." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).
Download Paper
CATS: Cascaded Adaptive Tree Speculation for Memory-Limited LLM Inference Acceleration
Published in arXiv preprint (in submission), 2026
A self-speculative decoding structure for memory-limited LLM inference on edge devices, achieving up to 5.08x wall-clock speedup with no degradation in generation quality, outperforming the prior SOTA method by up to 1.45x under edge memory constraints.
Recommended citation: Han, Y., Jin, Y., Zhao, D., & Sun, J. (2026). "CATS: Cascaded Adaptive Tree Speculation for Memory-Limited LLM Inference Acceleration." arXiv preprint arXiv:2605.11186.
Download Paper
LatentSift: Policy-State Filtering for Token-Efficient Verification of Software Engineering Agents
Published in arXiv preprint (in submission), 2026
A token-free, execution-free filter that verifies software engineering agents with the hidden states the policy already produced, cutting total verification tokens by 49.1–62.1% at K=16 while hybrid Best@16 matches or improves on each agent's reference workflow.
Recommended citation: Han, Y., Jin, Y., Jandreau, T., & Sun, J. (2026). "LatentSift: Policy-State Filtering for Token-Efficient Verification of Software Engineering Agents." arXiv preprint arXiv:2609.36371.
Download Paper