时间： 2023.4.17-2023.4.23
本周大事记
1. StableLM: Stability AI Language Models
StabilityAI开源了他们的大语言模型，目前3b和7b提供下载
2. 复旦大学开源大模型Moss
复旦 NLP 团队的 MOSS 大语言模型开源了，增加「搜索引擎、计算器、解方程、文生图」等插件功能，可以在线体验，支持本地部署
更多讨论：www.zhihu.com
最新技术：
Ask-Anything, tool for chatting about video with chatGPT, miniGPT4 and StableLM
github: github.com
Align your Latents: High-Resolution Video Synthesis with Latent Diffusion Models
论文：arxiv.org
Reference-based Image Composition with Sketch via Structure-aware Diffusion Model
论文：arxiv.org
Inpaint Anything: Segment Anything Meets Image Inpainting
github: github.com
demo: huggingface.co
DINOv2: Learning Robust Visual Features without Supervision
github: github.com
Anything-3D, present a project combining Segment Anything with a series of 3D models
github: github.com
LLM as A Robotic Brain: Unifying Egocentric Memory and Control
论文：arxiv.org
whisper-jax
github: github.com
Theory on Adam Instability in Large-Scale Machine Learning
论文: arxiv.org
Generative Disco: Text-to-Video Generation for Music Visualization
论文: arxiv.org
NaturalSpeech 2: Latent Diffusion Models are Natural and Zero-Shot Speech and Singing Synthesizers
论文: arxiv.org
SAM Fails to Segment Anything? -- SAM-Adapter: Adapting SAM in Underperformed Scenes: Camouflage, Shadow, and More
论文: arxiv.org
Avatars Grow Legs: Generating Smooth Human Motion from Sparse Tracking Inputs with Diffusion Model
论文: arxiv.org
Low-code LLM: Visual Programming over LLMs
论文: arxiv.org
Tool Learning with Foundation Models
论文: arxiv.org
github: github.com
Synthetic Data from Diffusion Models Improves ImageNet Classification
论文: arxiv.org
LongForm: Optimizing Instruction Tuning for Long Text Generation with Corpus Extraction
论文: arxiv.org
梯度视角下的LoRA：简介、分析、猜测及推
课程：
“生成式大语言模型技术分享”系列直播
如何生成文本: 通过 Transformers 用不同的解码方法生成文本
商业：
陆奇最新演讲实录：我的大模型世界观
王川: 从 chatGPT 看人工智能的投资机会和风险
对话王慧文：AGI这么伟大的事情，谁做成了我都会鼓掌
案例：
Chinese-LangChain
电商数字模特生成技术实践分享