AI研究論文デイリー

SAM 2: Segment Anything in Images and Videos

Gemma 2: Improving Open Language Models at a Practical Size

SF3D: Stable Fast 3D Mesh Reconstruction with UV-unwrapping and Illumination Disentanglement

Improving Text Embeddings for Smaller Language Models Using Contrastive Fine-tuning

OmniParser for Pure Vision Based GUI Agent

Coarse Correspondence Elicit 3D Spacetime Understanding in Multimodal Language Model

Finch: Prompt-guided Key-Value Cache Compression

TurboEdit: Text-Based Image Editing Using Few-Step Diffusion Models

MM-Vet v2: A Challenging Benchmark to Evaluate Large Multimodal Models for Integrated Capabilities

Reenact Anything: Semantic Video Motion Transfer Using Motion-Textual Inversion

UniTalker: Scaling up Audio-Driven 3D Facial Animation through A Unified Model

Tails Tell Tales: Chapter-Wide Manga Transcriptions with Character Names

Smoothed Energy Guidance: Guiding Diffusion Models with Reduced Energy Curvature of Attention

Non Verbis, Sed Rebus: Large Language Models are Weak Solvers of Italian Rebuses

Sentence-wise Speech Summarization: Task, Datasets, and End-to-End Modeling with LM Knowledge Distillation

Generalized Out-of-Distribution Detection and Beyond in Vision Language Model Era: A Survey

Enhancing Semantic Similarity Understanding in Arabic NLP with Nested Embedding Learning