Tech Papers, Training, and Development: Week Ending 09/13/2024
“Our latest course on LLM prompt evaluations is out. Evals ensure your prompts are production-ready as you’re able to quickly catch edge cases and zero in on exactly where your prompts need work. Let’s walk through what the course covers:
“I’ll be hosting a quick workshop on the AI Engineering Roadmap, aka skills to get hired as an ai engineer. Will be a free live-session with Q&A afterwards. I’d love to see you there, it’ll be next Sunday!
“GenAI implementation poses a series of hurdles to overcome. Choosing a vector database, data pre-processors, embeddings, deployment, and observability tools creates a complex puzzle. So, if I had a comprehensive toolkit to address the varied GenAI use cases and build compliant
“We’ve raised $12 million from Bessemer Venture Partners to build an AI-focused text editor that integrates tightly with our models.” / X
“I’m thrilled to announce the release of the digital version of “Hands-On Large Language Models” 🎉 The book contains more than 250 visuals (in color!) to help you understand the inner workings of LLMs and how to use them effectively.
“@leopd @karpathy The problem has to do with auto-regressive prediction, not with the architecture used to do so (transformers or whatever). Auto-regresssive prediction for things that are not temporal sequences (with some temporal causality) is a pure abomination. Even for temporal sequences,” / X
“Join us this week at AIAI Berlin for a session with Weights & Biases’ @hansramsl on “Generative AI in Manufacturing: Revolutionizing Tool Development.” Discover how top manufacturers use generative AI to innovate and optimize tool creation. Register here:
“Discover how to leverage the power of JavaScript and Python code steps within Relevance AI’s tool builder. In this tutorial, we cover: 🔗 Accessing step results and user inputs directly by variable names. ❓The differences between JavaScript and Python code steps. 📦 How to
“Chatbot Arena update! We added a new “Style Control” button to the leaderboard! Now you can apply it to Overall and Hard Prompts to see how rankings shift. We’re dedicated to continually improving the leaderboard, please share your feedback! Learn more details in our blog👇
“It is *incredibly* easy to game the LLM benchmarks. Training on test set is for the rookies. Here’re some tricks to practice magic at home: 1. Train on paraphrased examples of the test set. “LLM-decontaminator” paper from LMSys found that you can beat GPT-4 with a 13B model (!!)
“”Selective Reflection-Tuning” Paper 📚 The initial Reflection-Tuning approach of 2023 had some fundamental flaws. So an improved version called “Selective Reflection-Tuning” came (June-2024 ). This approach allows the student model to select which enhanced data from the teacher
[2409.04005v1] Qihoo-T2X: An Efficiency-Focused Diffusion Transformer via Proxy Tokens for Text-to-Any-Task
dailenson/One-DM: Official Code for ECCV 2024 paper — One-Shot Diffusion Mimicker for Handwritten Text Generation
2409.04109v1.pdf
chrome-extension://efaidnbmnnnibpcajpcglclefindmkaj/https://arxiv.org/pdf/2409.04109
“In-context learning (ICL) in LLMs, while powerful, is not fully understood. This new paper explicitly studies in-context learning in a regression problem and argues that ICL uses a combination of both learning from in-context examples and retrieving internal knowledge. They are
opal_ptx/notebooks/simple_tma.ipynb at master · kuterd/opal_ptx
[2409.01369] Imitating Language via Scalable Inverse Reinforcement Learning
“Good graph. NVDA still has no competition in training. It may have some competition on inference, but B100s with fp4 inference can blow the rest out of the picture very fast.
dmh2000/clone-layout
Leave a Reply