AI研究論文デイリー

Magpie: Alignment Data Synthesis from Scratch by Prompting Aligned LLMs with Nothing

NaRCan: Natural Refined Canonical Image with Integration of Diffusion Prior for Video Editing

What If We Recaption Billions of Web Images with LLaMA-3?

MotionClone: Training-Free Motion Cloning for Controllable Video Generation

Physics3D: Learning Physical Properties of 3D Gaussians via Video Diffusion

Are We Done with MMLU?

PowerInfer-2: Fast Large Language Model Inference on a Smartphone

VideoLLaMA 2: Advancing Spatial-Temporal Modeling and Audio Understanding in Video-LLMs

3D-GRAND: A Million-Scale Dataset for 3D-LLMs with Better Grounding and Less Hallucination

MMWorld: Towards Multi-discipline Multi-faceted World Model Evaluation in Videos

Turbo Sparse: Achieving LLM SOTA Performance with Minimal Activated Parameters

FontStudio: Shape-Adaptive Diffusion Model for Coherent and Consistent Font Effect Generation

Discovering Preference Optimization Algorithms with and for Large Language Models

AV-DiT: Efficient Audio-Visual Diffusion Transformer for Joint Audio and Video Generation

Hierarchical Patch Diffusion Models for High-Resolution Video Generation

Beyond LLaVA-HD: Diving into High-Resolution Large Multimodal Models

VCR: Visual Caption Restoration

Large Language Model Unlearning via Embedding-Corrupted Prompts

Chimera: Effectively Modeling Multivariate Time Series with 2-Dimensional State Space Models

Hibou: A Family of Foundational Vision Transformers for Pathology

Simplified and Generalized Masked Diffusion for Discrete Data