DGHMesh: The First Large-Scale Dual-Radar Millimeter-Wave Human Mesh Reconstruction Benchmark
A research team has released the DGHMesh dataset and benchmark, employing a dual millimeter-wave radar configuration to …
📂 Research
Latest articles in Research
WeatherSeg: A New Robust Image Segmentation Framework for Adverse Weather Conditions
A latest arXiv paper proposes WeatherSeg, a semi-supervised segmentation framework that significantly enhances environme…
SRMC: A Breakthrough in Constant-Memory Non-Markovian Samplers
Researchers propose the Score-Repellent Monte Carlo framework, which compresses trajectory history through sliding avera…
SGP-SAM: Self-Gated Prompting Mechanism Breaks Through 3D Medical Lesion Segmentation Bottleneck
A research team has proposed SGP-SAM, a method that leverages a self-gated prompting mechanism to address two major chal…
Behavioral Timescale Synaptic Plasticity: How a Single Experience Can Rewire the Brain
Cutting-edge neuroscience research reveals that Behavioral Timescale Synaptic Plasticity (BTSP) can rapidly rewire neura…
PivotMerge: A New Paradigm for Model Merging in Heterogeneous Multimodal Pretraining
A research team has proposed PivotMerge, a method that leverages post-alignment model merging techniques to effectively …
DO-Bench: A New Benchmark for Precisely Diagnosing Object Hallucination in Vision-Language Models
A research team has introduced the DO-Bench benchmark, which for the first time decomposes the causes of object hallucin…
FreqFormer: Frequency-Domain Attention Cracks the Efficiency Bottleneck of Long Video Generation
A latest arXiv paper proposes the FreqFormer framework, which dramatically reduces the computational cost of long-sequen…
PrivAR: Safeguarding AR Privacy with Vision-Language Models
A research team has proposed the PrivAR framework, which for the first time combines vision-language models with chain-o…
New Breakthrough in Offline Reinforcement Learning: Flexible Steering Even After Policy Freezing
A latest arXiv paper proposes a deploy-time adaptation framework for offline reinforcement learning based on Product-of-…
New Research Reveals Expressiveness Boundaries of Graph Neural Network Global Readout
A latest arXiv paper delves into the expressive power of message-passing graph neural networks (ACR-GNN), proving that s…
Study Reveals Self-Preference Bias in LLM Judges and Proposes Mitigation Strategies
A latest arXiv paper systematically quantifies the 'self-preference bias' phenomenon when large language models serve as…