Article

Explore our curated collection of AI articles covering artificial intelligence, machine learning, AI tools and more

263 Article•Page 1 / 30

ActionMesh: Animating 3D Mesh Generation with Temporal 3D Diffusion

ActionMesh: Animating 3D Mesh Generation with Temporal 3D Diffusion, capable of generating animated 3D meshes with fixed topology from input videos (real or synthetic).

Open source

COLIPRI: A 3D Vision-Language Transformer Model

COLIPRI is a 3D vision-language transformer model trained to encode chest CT scans and reports. It is trained on tens of thousands of chest CT scans and corresponding reports without any annotations, employing multi-task learning to develop a strong joint representation of 3D images and text.

Open source

PaddleOCR-VL-1.5, an Open-Source Document Parsing Model

PaddleOCR-VL-1.5: Towards a Multi-Task 0.9B VLM for Robust In-the-Wild Document Parsing

Open source

Qwen3-ASR: Qwen's Open-Source Speech Recognition Model

Qwen3-ASR is Qwen's open-source speech recognition model, leveraging large-scale speech training data and the powerful audio understanding capabilities of its foundational model, Qwen3-Omni.

Open source

Innovator-VL: A Scientific Multimodal Large Language Model

Innovator-VL, an open-source scientific multimodal large language model, advances multimodal understanding and reasoning across multiple scientific domains.

Open source

HunyuanImage-3.0-Instruct: Tencent's Open-Sourced Native Multimodal Image Generation Model

HunyuanImage-3.0-Instruct, a natively multimodal image generation model open-sourced by Tencent, features intelligent prompt enhancement and image-to-image generation with reasoning for creative editing.

Open source

Youtu-VL: Tencent's Open-Source Vision-Language Model

Youtu-VL, a vision-language model open-sourced by Tencent, adopts a novel VLUAS paradigm that treats vision as the objective rather than merely an input—enabling unified supervision without task-specific modules.

Open source

Kimi-K2.5: An Open-Source Native Multimodal Agent Model

Kimi-K2.5 is an open-source, native multimodal agent model built through continuous pre-training on approximately 15 trillion mixed visual and text tokens based on Kimi-K2-Base.