AI研究論文每日精選

API Agents vs. GUI Agents: Divergence and Convergence

PLADIS: Pushing the Limits of Attention in Diffusion Models at Inference Time by Leveraging Sparsity

MaRI: Material Retrieval Integration across Domains

Group-robust Machine Unlearning

CHOrD: Generation of Collision-Free, House-Scale, and Organized Digital Twins for 3D Indoor Scenes with Controllable Floor Plans and Optimal Layouts

Technologies on Effectiveness and Efficiency: A Survey of State Spaces Models

Neighboring Autoregressive Modeling for Efficient Visual Generation

Can Large Reasoning Models do Analogical Reasoning under Perceptual Uncertainty?

ProJudge: A Multi-Modal Multi-Discipline Benchmark and Instruction-Tuning Dataset for MLLM-based Process Judges

SmolDocling: An ultra-compact vision-language model for end-to-end multi-modal document conversion

Cockatiel: Ensembling Synthetic and Human Preferenced Training for Detailed Video Caption

TreeMeshGPT: Artistic Mesh Generation with Autoregressive Tree Sequencing

ARMOR v0.1: Empowering Autoregressive Multimodal Understanding Model with Interleaved Multimodal Generation via Asymmetric Synergy

ETCH: Generalizing Body Fitting to Clothed Humans via Equivariant Tightness

Kolmogorov-Arnold Attention: Is Learnable Attention Better For Vision Transformers?

TxAgent: An AI Agent for Therapeutic Reasoning Across a Universe of Tools

Learning Few-Step Diffusion Models by Trajectory Distribution Matching

Adversarial Data Collection: Human-Collaborative Perturbations for Efficient and Robust Robotic Imitation Learning

From TOWER to SPIRE: Adding the Speech Modality to a Text-Only LLM

Open-World Skill Discovery from Unsegmented Demonstrations

FlowTok: Flowing Seamlessly Across Text and Image Tokens

Vamba: Understanding Hour-Long Videos with Hybrid Mamba-Transformers

GoalFlow: Goal-Driven Flow Matching for Multimodal Trajectories Generation in End-to-End Autonomous Driving

ReCamMaster: Camera-Controlled Generative Rendering from A Single Video

Exploring the Vulnerabilities of Federated Learning: A Deep Dive into Gradient Inversion Attacks

Large-scale Pre-training for Grounded Video Caption Generation

VGGT: Visual Geometry Grounded Transformer