KI-Forschungspapiere Täglich

ReCamMaster: Camera-Controlled Generative Rendering from A Single Video

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

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

API Agents vs. GUI Agents: Divergence and Convergence

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

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

VGGT: Visual Geometry Grounded Transformer

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

FlowTok: Flowing Seamlessly Across Text and Image Tokens

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

Large-scale Pre-training for Grounded Video Caption Generation

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

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

ETCH: Generalizing Body Fitting to Clothed Humans via Equivariant Tightness

Neighboring Autoregressive Modeling for Efficient Visual Generation

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

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

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

Learning Few-Step Diffusion Models by Trajectory Distribution Matching

TreeMeshGPT: Artistic Mesh Generation with Autoregressive Tree Sequencing

MaRI: Material Retrieval Integration across Domains

Can Large Reasoning Models do Analogical Reasoning under Perceptual Uncertainty?

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

Open-World Skill Discovery from Unsegmented Demonstrations

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

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

Group-robust Machine Unlearning