5 papers
BeyondFacial: Identity-Preserving Personalized Generation Beyond Facial Close-ups
Songsong Zhang, Chuanqi Tang, Hongguang Zhang +7
Identity-Preserving Personalized Generation (IPPG) has advanced film production and artistic creation, yet existing approaches overemphasize facial regions, resulting in outputs do…
ZSPAPrune: Zero-Shot Prompt-Aware Token Pruning for Vision-Language Models
Pu Zhang, Yuwei Li, Xingyuan Xian +1
As the capabilities of Vision-Language Models (VLMs) advance, they can process increasingly large inputs, which, unlike in LLMs, generates significant visual token redundancy and l…
Characterizing Model Behavior Under Synthetic Data Training: An Empirical Study Across Scales and Mixing Ratios
Y. Du, G. Wu, G. Tang +2
Synthetic data generated by large language models has become integral to modern NLP training pipelines, from bootstrapping reasoning capabilities to augmenting instruction-followin…
Cognitive Decision Routing in Large Language Models: When to Think Fast, When to Think Slow
Y. Du, C. Guo, W. Wang +1
Large Language Models (LLMs) face a fundamental challenge in deciding when to rely on rapid, intuitive responses versus engaging in slower, more deliberate reasoning. Inspired by D…
MRCBert: A Machine Reading ComprehensionApproach for Unsupervised Summarization
Saurabh Jain, Guokai Tang, Lim Sze Chi
When making an online purchase, it becomes important for the customer to read the product reviews carefully and make a decision based on that. However, reviews can be lengthy, may…