开源项目
最新开源 AI 模型、工具、框架与社区项目推荐。
Tucano 2 Cool: Better Open Source LLMs for Portuguese
Tucano 2 is a comprehensive, fully open-source suite of large language models specifically engineered for the Portuguese...
Tucano 2 Cool: Better Open Source LLMs for Portuguese
Tucano 2 is a suite of open-source large language models specifically optimized for Portuguese, serving over 260 million...
Tucano 2 Cool: Better Open Source LLMs for Portuguese
Tucano 2 is a comprehensive open-source suite of large language models specifically engineered for Portuguese, featuring...
Tucano 2 Cool: Better Open Source LLMs for Portuguese
Tucano 2 is a suite of open-source large language models specifically optimized for Portuguese, addressing the scarcity ...
Tucano 2 Cool: Better Open Source LLMs for Portuguese
Tucano 2 is a suite of fully open-source large language models (0.5B to 3.7B parameters) specifically optimized for Port...
mlx-snn: Spiking Neural Networks on Apple Silicon via MLX
mlx-snn is the first native spiking neural network library for Apple's MLX framework, enabling efficient neuromorphic co...
mlx-snn: Spiking Neural Networks on Apple Silicon via MLX
mlx-snn is the first dedicated spiking neural network library built natively for Apple's MLX framework, enabling efficie...
mlx-snn: Spiking Neural Networks on Apple Silicon via MLX
mlx-snn is the first dedicated spiking neural network library built natively for Apple's MLX framework, optimized for Ap...
mlx-snn: Spiking Neural Networks on Apple Silicon via MLX
mlx-snn is the inaugural spiking neural network library built natively on Apple's MLX framework for Apple Silicon. It ac...
Bridging the Reproducibility Divide: Open Source Software's Role in Standardizing Healthcare AI
Healthcare AI research faces a reproducibility crisis, with 74% of studies relying on private data or unshared code, pre...
How our open-source AI model SpeciesNet is helping to promote wildlife conservation
Google has launched SpeciesNet, an open-source AI model specifically designed for wildlife identification and conservati...
千问模型负责人林俊旸提出离职，阿里高管紧急答疑 | 智能涌现独家
2025年3月，阿里巴巴通义千问（Qwen）大模型技术负责人林俊旸突然宣布离职，引发团队多名核心成员相继提出离职，在AI社区与阿里内部造成巨大震动。阿里集团CEO吴泳铭等高层紧急召开会议，否认团队收缩，强调此次为旨在扩充人才与资源的扩张性调...
千问模型负责人林俊旸提出离职，阿里高管紧急答疑 | 智能涌现独家
阿里巴巴通义千问（Qwen）大模型技术负责人林俊旸于3月4日突然宣布离职，引发团队核心成员跟随离职的连锁反应。阿里高层紧急召开会议，将此次组织调整定性为团队“扩张”而非“收缩”，并强调千问基础模型研发是当前“阿里巴巴集团层面最重要的事项”。...
智元灵渠OS开源上线
智元机器人正式开源其灵渠OS Alpha版本，这是一个基于已量产的全尺寸人形机器人“远征A2”本体开发的全栈机器人操作系统。开源内容包括跨平台具身软件框架、基于强化学习的双足运动控制框架，以及一站式仿真训练部署工具链，旨在降低行业技术门槛并...
摩尔线程MTT S5000全面适配阿里Qwen3.5三款新模型
36氪获悉，2月26日，摩尔线程官微宣布已在AI训推一体全功能GPU MTT S5000上，完成对阿里三款全新模型的全方位适配。据介绍，继开源Qwen3.5-397B-A17B之后，阿里宣布开源千问3.5最新三款中等规模模型Qwen3.5-...
Hyperbolic Busemann Neural Networks
arXiv:2602.18858v2 Announce Type: replace-cross Abstract: Hyperbolic spaces provide a natural geometry for representing ...
Decoding as Optimisation on the Probability Simplex: From Top-K to Top-P (Nucleus) to Best-of-K Samplers
arXiv:2602.18292v2 Announce Type: replace-cross Abstract: Decoding sits between a language model and everything we do wi...
Diffusion Language Models Know the Answer Before Decoding
arXiv:2508.19982v4 Announce Type: replace-cross Abstract: Diffusion language models (DLMs) have recently emerged as an a...
Modular Deep Learning for Multivariate Time-Series: Decoupling Imputation and Downstream Tasks
arXiv:2411.03941v3 Announce Type: replace-cross Abstract: Missing values are pervasive in large-scale time-series data, ...
Spilled Energy in Large Language Models
arXiv:2602.18671v2 Announce Type: replace Abstract: We reinterpret the final Large Language Model (LLM) softmax classifi...
Spurious Rewards: Rethinking Training Signals in RLVR
arXiv:2506.10947v2 Announce Type: replace Abstract: We show that reinforcement learning with verifiable rewards (RLVR) c...
Temporal Knowledge-Graph Memory in a Partially Observable Environment
arXiv:2408.05861v4 Announce Type: replace Abstract: Agents in partially observable environments require persistent memor...
Off-The-Shelf Image-to-Image Models Are All You Need To Defeat Image Protection Schemes
arXiv:2602.22197v1 Announce Type: cross Abstract: Advances in Generative AI (GenAI) have led to the development of vario...
On Imbalanced Regression with Hoeffding Trees
arXiv:2602.22101v1 Announce Type: cross Abstract: Many real-world applications provide a continuous stream of data that ...
DualWeaver: Synergistic Feature Weaving Surrogates for Multivariate Forecasting with Univariate Time Series Foundation Models
arXiv:2602.22066v1 Announce Type: cross Abstract: Time-series foundation models (TSFMs) have achieved strong univariate ...
RGB-Event HyperGraph Prompt for Kilometer Marker Recognition based on Pre-trained Foundation Models
arXiv:2602.22026v1 Announce Type: cross Abstract: Metro trains often operate in highly complex environments, characteriz...
xai-cola: A Python library for sparsifying counterfactual explanations
arXiv:2602.21845v1 Announce Type: cross Abstract: Counterfactual explanation (CE) is an important domain within post-hoc...
Learning from Yesterday's Error: An Efficient Online Learning Method for Traffic Demand Prediction
arXiv:2602.21757v1 Announce Type: cross Abstract: Accurately predicting short-term traffic demand is critical for intell...
ECHOSAT: Estimating Canopy Height Over Space And Time
arXiv:2602.21421v1 Announce Type: cross Abstract: Forest monitoring is critical for climate change mitigation. However, ...
Small Language Models for Privacy-Preserving Clinical Information Extraction in Low-Resource Languages
arXiv:2602.21374v1 Announce Type: cross Abstract: Extracting clinical information from medical transcripts in low-resour...
又快又省？仅5%参数、训练快4倍！ArcFlow用「非线性」魔法实现FLUX/Qwen推理40倍加速
在生成式 AI 的浪潮中，我们见证了从 Stable Diffusion 到 FLUX、Qwen-Image 等大规模扩散模型的画质飞跃。然而，这种飞跃并非没有代价。为了从纯噪声中 “雕刻” 出清晰的图像，这些模型通...
DeepMind药物衍生公司的独家新AI，堪称AlphaFold 4的专有药物设计引擎
编辑丨&在谷歌 DeepMind 发布了针对药物发现的更新版 AlphaFold3 近两年后，其生物制药衍生公司 Isomorphic Labs 宣布了更强大的人工智能模型—&md...
最强Coding Plan上线！阿里云上线Qwen3.5、GLM-5、MiniMax M2.5、Kimi K2.5四大顶尖开源模型
2月25日，阿里云百炼推出包含Qwen3.5、GLM-5、MiniMax M2.5、Kimi K2.5四大顶尖开源模型API服务的最强Coding Plan。用户订阅套餐后不再受限于单一模型，可实现多模型自由切换，享受更稳定、Tokens额...
消费级显卡可跑！刚刚，阿里Qwen3.5又开源3款新模型
刚过完年，阿里又卷起来了。2 月 25 日，继除夕开源 Qwen3.5-397B-A17B 之后，阿里继续开源千问 3.5 系列模型，而且是一口气开源三款中等规模的新模型，分别是 Qwen3.5-35B-A3B、Qwen3.5-1...
PMG: Parameterized Motion Generator for Human-like Locomotion Control
arXiv:2602.12656v2 Announce Type: replace-cross Abstract: Recent advances in data-driven reinforcement learning and moti...
UI-Venus-1.5 Technical Report
arXiv:2602.09082v2 Announce Type: replace-cross Abstract: GUI agents have emerged as a powerful paradigm for automating ...
GOT-Edit: Geometry-Aware Generic Object Tracking via Online Model Editing
arXiv:2602.08550v3 Announce Type: replace-cross Abstract: Human perception for effective object tracking in 2D video str...
MoMaGen: Generating Demonstrations under Soft and Hard Constraints for Multi-Step Bimanual Mobile Manipulation
arXiv:2510.18316v3 Announce Type: replace-cross Abstract: Imitation learning from large-scale, diverse human demonstrati...
LD-MoLE: Learnable Dynamic Routing for Mixture of LoRA Experts
arXiv:2509.25684v2 Announce Type: replace-cross Abstract: Recent studies have shown that combining parameter-efficient f...
CONTINA: Confidence Interval for Traffic Demand Prediction with Coverage Guarantee
arXiv:2504.13961v2 Announce Type: replace-cross Abstract: Accurate short-term traffic demand prediction is critical for ...
Hidden Dynamics of Massive Activations in Transformer Training
arXiv:2508.03616v2 Announce Type: replace Abstract: We present the first comprehensive analysis of massive activation de...
Efficient Hierarchical Any-Angle Path Planning on Multi-Resolution 3D Grids
arXiv:2602.21174v1 Announce Type: cross Abstract: Hierarchical, multi-resolution volumetric mapping approaches are widel...
SparkMe: Adaptive Semi-Structured Interviewing for Qualitative Insight Discovery
arXiv:2602.21136v1 Announce Type: cross Abstract: Qualitative insights from user experiences are critical for informing ...
MIP Candy: A Modular PyTorch Framework for Medical Image Processing
arXiv:2602.21033v1 Announce Type: cross Abstract: Medical image processing demands specialized software that handles hig...
Dataset Color Quantization: A Training-Oriented Framework for Dataset-Level Compression
arXiv:2602.20650v1 Announce Type: cross Abstract: Large-scale image datasets are fundamental to deep learning, but their...
POMDPPlanners: Open-Source Package for POMDP Planning
arXiv:2602.20810v1 Announce Type: new Abstract: We present POMDPPlanners, an open-source Python package for empirical ev...
Train AI models with Unsloth and Hugging Face Jobs for FREE
GGML and llama.cpp join HF to ensure the long-term progress of Local AI
Deploying Open Source Vision Language Models (VLM) on Jetson
Spanish ‘soonicorn’ Multiverse Computing releases free compressed AI model
Spanish startup Multiverse Computing has released a new version of its HyperNova 60B model on Hugging Face that, it says...