For the first time, we’ve extracted millions of features from a high-performing, deployed model (Claude 3 Sonnet). These features cover specific people and places, programming-related abstractions, scientific topics, emotions, among a vast range of other concepts.
MACHINE LEARNING
-
Dictionary Learning Makes LLM Neurons More Interpretable
By
–
The problem: most LLM neurons are uninterpretable, stopping us from mechanistically understanding the models. In October, we showed that dictionary learning could decompose a small model into "monosemantic" components we call "features"—making the model more interpretable.
-
Mastering Recommendation Systems: A Complete Guide
By
–
Ever wondered how Spotify, YouTube, Netflix, or Amazon know just what you like? It's all thanks to recommendation systems!https://t.co/HSt9hBJSt8
— Satya Mallick (@LearnOpenCV) 21 mai 2024
Our latest blog "Mastering Recommendation Systems: A Complete Guide" breaks down what a recommendation system is and how it works.… pic.twitter.com/dAPNxzXlhfEver wondered how Spotify, YouTube, Netflix, or Amazon know just what you like? It's all thanks to recommendation systems! https://
learnopencv.com/recommendation
-system/
… Our latest blog "Mastering Recommendation Systems: A Complete Guide" breaks down what a recommendation system is and how it works. -
RAG Knowledge Graphs: Detailed Technical Walkthrough
By
–
detailed walkthrough of how to RAG against knowledge graphs
-
YOLOv5s 16-Stream Demo: Fast AI Inference Deployment Kit
By
–
Check out our 16-stream demo using YOLOv5s for object recognition. Deployment takes just 10 minutes, and no retraining is needed. Get your Metis Evaluation Kit now to explore AI inference technology and accelerate your innovation. https://t.co/XYy5NcavaM#AI #MachineLearning pic.twitter.com/0TP9mNMWGQ
— Axelera AI (@AxeleraAI) 21 mai 2024Check out our 16-stream demo using YOLOv5s for object recognition. Deployment takes just 10 minutes, and no retraining is needed. Get your Metis Evaluation Kit now to explore AI inference technology and accelerate your innovation. https://
tinyurl.com/244md3rf
#AI #MachineLearning -
Balancing AI Automation with Human Empathy and Ethics
By
–
The Human Side of #AI: Insights on Balancing #Automation and #Empathy https://
innotechtoday.com/the-human-side
-of-ai-insights-on-balancing-automation-and-empathy/
… @EnricoPalmerino #MachineLearning @enilev @mvollmer1 @CurieuxExplorer @segundoatdell @lyakovet @HeinzVHoenen @Eli_Krumova @CatherineAdenle @Damien_CABADI @FmFrancois @AkwyZ @avrohomg -
Neural Pathways: Connections and Sparsity in Network Training
By
–
It's an homage to neural pathways, which seemed appropriate since neural network training is all about making connections between neurons (and also coveys sparsity, which we were/are interested in). "Neural pathways are the connections between neurons that light up when you
-
Physics-ML in Education: NVIDIA Modulus Webinar for Engineers
By
–
Discover the power of physics-ML in modeling real-world systems, and explore its application in #education in this webinar. Empower future engineers and scientists with NVIDIA Modulus. Register now: https://
nvda.ws/4axVBvo -
Running Evals: Scripts, Model Swapping Libraries, and Tools
By
–
how do you run these evals? is it just a bunch of scripts? do you use any libraries for swapping models? any other libraries useful in your eval?
-
Prompting trick: split example responses into half-messages
By
–
This won’t help in zero shot but if you have in-context examples try splitting each example response into two half-messages, so it appears assistant responses are expected come in pairs with a semantically meaningless split.