Qwen2.5-1M is an open-source model released by the Alibaba Tongyi Qianwen team, supporting context lengths of 1 million tokens. The model includes two versions: Qwen2.5-7B-Instruct-1M and Qwen2.5-14B-Instruct-1M. (The text abruptly ends here, likely due to an incomplete sentence or missing information.)
DiffuEraser is a video inpainting model based on a stable diffusion model that fills masked areas in videos with richer detail and more coherent structure. The model provides initialization and weak conditions by incorporating prior information, which helps reduce...
Janus-Pro is an open-source AI model from DeepSeek that supports image understanding and generation, offering both 1B and 7B scales to suit diverse application scenarios. It features improved training strategies, expanded datasets, and larger scale...
Qwen2.5-VL is the flagship visual language model open-sourced by Alibaba's Tongyi Qianwen team, available in three different scales: 3B, 7B, and 72B. The model excels in visual understanding, capable of recognizing common objects and analyzing text and charts in images...
QVQ-72B-Preview is an open-source multimodal reasoning model from Alibaba Cloud's Tongyi Qianwen team, focusing on improving visual reasoning capabilities. The model performs exceptionally well in multiple benchmark tests, demonstrating powerful capabilities in multimodal understanding and reasoning tasks...
Qwen2.5-Max is a large-scale MoE (Mixture of Experts) model launched by Alibaba Cloud, using over 20 trillion pre-trained tokens. The model performs exceptionally well in multiple benchmark tests, surpassing DeepSeek V3, ...
MobileVD (Mobile Video Diffusion) is the first video diffusion model optimized for mobile devices, developed by the Qualcomm AI Research team. The model is based on the spatiotemporal UN structure of Stable Video Diffusion (SVD)...
YuE is an open-source AI music generation model jointly developed by the Hong Kong University of Science and Technology and the Multimodal Art Projection team. It can convert lyrics into complete songs and supports various music styles, including pop, metal, jazz, etc.
VARGPT is an innovative multimodal large language model focused on visual understanding and generation tasks. Based on an autoregressive framework, it unifies visual generation and understanding within a single model, avoiding the complexity of task switching. VARGPT is built on the LLaVA architecture...
CityDreamer4D is a composite generative model developed by the S-Lab team at Nanyang Technological University for generating borderless 4D cities. It separates dynamic objects (such as vehicles) from static scenes (such as buildings and roads) through three modules—...
SpeechGPT 2.0-preview is a human-like real-time interactive system developed by the OpenMOSS team at Fudan University. Trained on millions of hours of Chinese speech data, it employs an end-to-end architecture and achieves a high degree of integration between speech and text modalities...
LalaEval is a human evaluation framework for domain-specific Large Language Models (LLMs), jointly developed by the Chinese University of Hong Kong and the Lalamove Data Science team. The framework utilizes a complete end-to-end protocol, covering domain specifications, standards, etc.
Aligner is a large language model alignment technique proposed by a team from Peking University. It improves model performance by learning the corrected residuals between aligned and unaligned answers. It employs an autoregressive seq2seq model, working in the question-answer-corrected residual relationship...
OpenAI o3-mini is a brand-new inference model officially released by OpenAI. It is the latest and most cost-effective model in the OpenAI inference series, optimized for scientific, mathematical, and programming technologies. It supports three inference intensities (low...
EICopilot is an AI-powered enterprise information search and exploration tool developed by Baidu Research. Through an AI agent driven by a large language model (LLM), it helps enterprise users efficiently retrieve and interpret information within large-scale knowledge graphs...
Mistral Small 3 is an open-source large-scale language model released by Mistral AI, containing 24 billion parameters and licensed under the Apache 2.0 license. It is designed for low-latency tasks and supports multiple languages (including English, Chinese, Japanese, etc.).
CoA (Chain-of-Agents) is a multi-agent collaborative framework launched by Google to address the contextual constraints faced by Large Language Models (LLMs) when processing long text tasks. The CoA framework segments long texts into multiple shorter...
Oumi is a fully open-source AI platform that simplifies the entire lifecycle from data preparation and model training to evaluation and deployment. It supports model training with 10 million to 405 billion parameters, covering text and multimodal models (such as Lla...).
IC-Portrait is a novel personalized portrait generation framework that addresses the challenges posed by the diversity of user profile images (such as differences in appearance and lighting conditions). It breaks down the portrait generation task into "light-aware stitching" and "viewpoint consistency..."
SyncAnimation is a real-time, end-to-end audio-driven framework for generating human posture and speaking avatar animations. It generates upper body postures and facial expressions, including lip movements, synchronized with the audio signal in real time, achieving high-fidelity animation...
BitsAI-CR is an automated code review tool developed by ByteDance based on a large language model (LLM). It achieves efficient code review through a two-stage process: RuleChecker is responsible for detecting potential code errors based on 219 rules...
DiffSplat is a novel 3D generation method that rapidly generates 3D Gaussian splats from text prompts and single-view images. It leverages a pre-trained text-to-image diffusion model, based on a powerful 2D prior...
X-Prompt is a general framework for multimodal video object segmentation, addressing the limitations of traditional methods in complex scenarios such as extreme lighting, fast motion, and background interference. It works by pre-training a video object segmentation algorithm based on RGB data...
VideoReward is a video generation preference dataset and reward model jointly created by the Chinese University of Hong Kong, Tsinghua University, and Kuaishou Technology. It contains 182,000 labeled data points, covering three aspects: visual quality, motion quality, and text alignment...