AI论文简报
搜索
方法论
公众号
EN
千段视频接通4D,六成头更新可删
从331篇论文中选出11篇
重点关注
VibeLifeBench: Can Your Life Agent Be Proactive and Persistent in a Living World?
score 9
入选 HF Daily Papers;HF 热度: 17 upvotes (+3);有代码实现;关键词(1): open-source
也值得关注
SPIEval: Evaluating Large Language Models as Mobile Assistants over Scattered Personal Information
score 7
入选 HF Daily Papers;HF 热度: 11 upvotes (+3);关键词(1): reasoning
Flow Straight to Reality: Perceptually Consistent Flow Matching for Efficient Image Restoration
score 4
关键词(1): lightweight;顶会接收: ECCV
UniProbe: A Learnable Token-Level Hallucination Detector for Large VLMs using Multi-Structural Internal Representations
score 4
机构: NVIDIA;关键词(5): lightweight, latency, fine-tuning, reasoning, vision-language
Diffract: Spectral View of LLM Domain Adaptation
score 4
关键词(2): pre-training, open-source;顶会接收: ICML
Every Packet Counts: Dispersing Information for Loss-Resilient Learned Image Compression
score 4
机构: Huawei;关键词(2): compression, coding
A second-order theory of texture for depth from focus
score 3
机构: CMU
DSAR: Dual-Stream Autoregressive Modeling of Temporal Cloth Dynamics for Photorealistic Animatable Avatars
score 3
顶会接收: ECCV
Beyond Pixels: From Video Priors to 4D Worlds
score 9
入选 HF Daily Papers;HF 热度: 171 upvotes (+4);有代码实现
Beyond Fixed Luminance: Towards Panchromatic and Orthochromatic Image Colorization
score 3
顶会接收: ECCV
HNDiff: Haze-Noise Diffusion for Image Dehazing
score 3
顶会接收: ECCV