Runpeng Yu

Every Step Counts: Decoding Trajectories as Authorship Fingerprints of dLLMs

Dimple: Discrete Diffusion Multimodal Large Language Model with Parallel Decoding

NoLan: Mitigating Object Hallucinations in Large Vision-Language Models via Dynamic Suppression of Language Priors

Vid-SME: Membership Inference Attacks against Large Video Understanding Models

HG-Adapter: Improving Pre-Trained Heterogeneous Graph Neural Networks with Dual Adapters

dKV-Cache: The Cache for Diffusion Language Models

Neural Phylogeny: Fine-Tuning Relationship Detection among Neural Networks

Injecting Vision Language into Autoregressive Image Generation