时间： 2023.5.22-2023.5.28
本周大事记
1. Meta发布大规模多语言语音(MMS)项目
2. State of GPT：Andrej Karpathy揭秘OpenAI大模型原理和训练过程
视频：
ppt： https://karpathy.ai/stateofgpt.pdf
最新技术：
Role-Play with Large Language Models
论文：arxiv.org
PandaGPT: One Model To Instruction-Follow Them All
论文：arxiv.org
Impossible Distillation: from Low-Quality Model to High-Quality Dataset & Model for Summarization and Paraphrasing
论文：arxiv.org
Backpack Language Models
论文：arxiv.org
Training Socially Aligned Language Models in Simulated Human Society
论文：arxiv.org
代码：github.com
OlaGPT: Empowering LLMs With Human-like Problem-Solving Abilities
论文：arxiv.org
Break-A-Scene: Extracting Multiple Concepts from a Single Image
论文：arxiv.org
Voyager: An Open-Ended Embodied Agent with Large Language Models
论文：arxiv.org
代码：github.com
ProlificDreamer: High-Fidelity and Diverse Text-to-3D Generation with Variational Score Distillation
论文：arxiv.org
On Architectural Compression of Text-to-Image Diffusion Models
论文：arxiv.org
Unsupervised Semantic Correspondence Using Stable Diffusion
论文：arxiv.org
The False Promise of Imitating Proprietary LLMs
论文：arxiv.org
PEARL: Prompting Large Language Models to Plan and Execute Actions Over Long Documents
论文：arxiv.org
Enhancing Detail Preservation for Customized Text-to-Image Generation: A Regularization-Free Approach
论文：arxiv.org
Control-A-Video: Controllable Text-to-Video Generation with Diffusion Models
ControlVideo: Training-free Controllable Text-to-Video Generation
论文：arxiv.org
RWKV: Reinventing RNNs for the Transformer Era
Textually Pretrained Speech Language Models
论文：arxiv.org
Training Diffusion Models with Reinforcement Learning
论文：arxiv.org
Comparing Software Developers with ChatGPT: An Empirical Investigation
论文：arxiv.org
Controlling the Extraction of Memorized Data from Large Language Models via Prompt-Tuning
论文：arxiv.org
Pengi: An Audio Language Model for Audio Tasks
论文：arxiv.org
Cross-Lingual Supervision improves Large Language Models Pre-training
论文：arxiv.org
Optimizing Stable Diffusion for Intel CPUs with NNCF and Optimum
世界的参数倒影：为何GPT通过Next Token Prediction可以产生智能
课程：
如何使用 Megatron-LM 训练语言模型
斯坦福CS25课程：Transformers United V2
AI Canon:生成人工智能资源清单
商业：
Chatbot Arena Leaderboard Updates (Week 4)
大模型LLM领域，有哪些可以作为学术研究方向？
中国的OpenAI，藏在这几栋楼里
“AI教父”Geoffrey Hinton：智能进化的下一个阶段
C-Eval: 构造中文大模型的知识评估基准
创业公司的九种商业模式和定价策略
案例：
类DragGAN演示视频
暴雪自研AIGC图像工具Blizzard Diffusion：部分美术团队已在试用mp.weixin.qq.com
AI大模型如何在行业实际落地：企业对话场景拥抱大模型之路
翻译视频并对齐嘴型