5 papers
RUTA: Principled Visual Token Allocation via Rate-Utility Optimization
Jian Zou, Xiaoyu Xu, Zhihua Wang +3
High-resolution images and long videos provide vision-language models with rich context for multimodal reasoning and fine-grained perception, but the resulting long visual token se…
Learning Stable Canonical Worlds for Novel View Synthesis and Beyond
Xiaoyu Xu, Jian Zou, Sheyang Tang +3
Feed-forward Gaussian splatting (FFGS) facilitates real-time novel view synthesis, yet current methods often remain tied to view-dependent predictions. As more input views are adde…
MDS-VQA: Model-Informed Data Selection for Video Quality Assessment
Jian Zou, Xiaoyu Xu, Zhihua Wang +3
Learning-based video quality assessment (VQA) has advanced rapidly, yet progress is increasingly constrained by a disconnect between model design and dataset curation. Model-centri…
Evaluating Reward Model Generalization via Pairwise Maximum Discrepancy Competitions
Shunyang Luo, Peibei Cao, Zhihui Zhu +3
Reward models (RMs) are central to aligning large language models, yet their practical effectiveness hinges on generalization to unseen prompts and shifting distributions. Most exi…
Sample-Efficient Human Evaluation of Large Language Models via Maximum Discrepancy Competition
Kehua Feng, Keyan Ding, Hongzhi Tan +8
Reliable evaluation of large language models (LLMs) is impeded by two key challenges: objective metrics often fail to reflect human perception of natural language, and exhaustive h…