arXivDaily arXiv每日学术速递 周一至周五更新

高校专区

The Hong Kong University of Science and Technology(香港科技大学)

2026-08-03 至 2026-08-03 共收录 3
2607.23125 2026-08-03 cs.LG 版本更新

Self-Boosting Vision-Language Models with Noisy Student On-Policy Self-Distillation

基于有噪学生策略性自蒸馏的自增强视觉语言模型

Shuai Wang, Daoan Zhang, Zhe Tang, Hao Cheng, Jiaheng Wei

机构 * The Hong Kong University of Science and Technology (Guangzhou)(香港科技大学(广州)) ByteDance Inc.(字节跳动公司) Zhejiang University of Technology(浙江工业大学) Hong Kong Baptist University(香港浸会大学)

AI总结 研究针对视觉语言模型无外部监督难以改进的问题,提出NOPD自蒸馏方法,利用干净与损坏输入预测差异产生自监督信号,在多视觉推理任务中有效,能提升多个模型在多个基准测试中的表现。

详情

展开后加载摘要…

URL PDF HTML 收藏
2602.10429 2026-08-03 cs.MA cs.AI 版本更新

AIvilization v0: Toward Large-Scale Artificial Social Simulation with a Unified Agent Architecture and Adaptive Agent Profiles

AIvilization v0:迈向大规模人工社会模拟的统一代理架构与自适应代理档案

Wenkai Fan, Shurui Zhang, Xiaolong Wang, Haowei Yang, Tsz Wai Chan, Xingyan Chen, Junquan Bi, Zirui Zhou, Jia Liu, Kani Chen

机构 * The Hong Kong University of Science and Technology(香港科学与技术大学) Bauhinia AI

AI总结 AIvilization v0通过统一代理架构和自适应代理档案,实现大规模人工社会模拟,解决目标稳定性与反应正确性矛盾,支持长周期自主性和多目标环境下的稳健性。

Comments v2: major revision. Agent architecture and environment consolidated into self-contained sections; new problem-setting section formalizing the asynchronous event model; evaluation reorganized by experiment with results reported alongside each protocol; added action-simulator audit, and human-steering analyses; other sections rewritten

详情

展开后加载摘要…

URL PDF HTML 收藏
2505.18660 2026-08-03 cs.CV 版本更新

So-Fake: Benchmarking and Explaining Social Media Image Forgery Detection

So-Fake:社交媒体图像伪造检测的基准构建与可解释性研究

Zhenglin Huang, Xiangtai Li, Xi Yang, Bei Peng, Xiaowei Huang, Baoyuan Wu, Dacheng Tao, Ming-Hsuan Yang, Guangliang Cheng

机构 * University of Liverpool, UK(利物浦大学) Nanyang Technological University(南洋理工大学) Zhejiang University(浙江大学) National University of Singapore(新加坡国立大学) The Hong Kong University of Science and Technology(香港科学与技术大学) The Chinese University of Hong Kong, Shenzhen(香港中文大学(深圳)) The University of Sheffield(谢菲尔德大学)

AI总结 该研究针对社交媒体图像伪造检测的数据集与基准缺口,构建了So-Fake-Set数据集与So-Fake-OOD分布外基准,提出采用强化学习的So-Fake-R1框架,其检测准确率与定位IoU均优于现有方法,为相关研究提供了新基础。

详情

展开后加载摘要…

URL PDF HTML 收藏