CORE LLM

Interfaces for exploring transformer language models by looking at input saliency and neuron activation. Explorable #1: Input saliency of a list of countries generated by a language model Tap or hover over the output tokens: Explorable #2: Neuron activation analysis reveals four groups of neurons, each is associated with generating a certain type of token Tap or hover over the sparklines on the left to isolate a certain factor: The Transformer architecture has been powering a number of the recent advances in NLP. A breakdown of this architecture is provided here . Pre-trained language models based on the architecture, in both its auto-regressive (models that use their own output as input to next time-steps and that process tokens from left-to-right, like GPT2) and denoising (models trained by corrupting/masking the input and that process tokens bidirectionally, like BERT) variants continue to push the envelope in various tasks in NLP and, more recently, in computer vision. Our understanding of why these models work so well, however, still lags behind these developments. This exposition series continues the pursuit to interpret and visualize the inner-workings of transformer-based language models. We illustrate how some key interpretability methods apply to transformer-based language models. This article focuses on auto-regressive models, but these methods are applicable to other architectures and tasks as well. This is the first article in the series. In it, we present explorables and visualizations aiding the intuition of: Input Saliency methods that score input tokens importance to generating a token. Neuron Activations and how individual and groups of model neurons spike in response to inputs and to produce outputs. The next article addresses Hidden State Evolution across the layers of the model and what it may tell us about each layer's role.

Tokenization is a foundational process used across various fields of AI, primarily in natural language processing (NLP), It involves breaking down raw data into smaller, manageable units called "tokens" that AI models can process numerically. [1, 2, 3, 4, 5] Key areas where tokenization is used ...

Foundation models are large AI models trained on massive, diverse datasets that can be adapted to a wide range of downstream tasks. Instead of being built for one specific purpose, like traditional machine learning models, they serve as a flexible base for many applications, such as natural lan...

For your inspiration, read later, media and stuff

The primary uses of NLP include automating tasks like data analysis, machine translation, and customer service through chatbots. It also powers everyday applications such as speech recognition for voice assistants, email spam filters, and text summarization. NLP helps computers understand and pro...

NLP is intrinsically linked with Transformer models. The Transformer is a revolutionary deep learning architecture that is now a foundation for most modern NLP tasks, including machine translation, text generation, and summarization. Transformers are used to process and understand human l...

GD

GD

Referenced Embeddings are numerical representations of high-dimensional data like text and images, transformed into lower-dimensional vectors that machine learning models can process efficiently. These vectors capture semantic relationships, so similar items are placed closer together in the em...

AI hidden states are vectors that represent the intermediate memory of a neural network, particularly recurrent neural networks (RNNs) and Transformers. In an RNN, the hidden state vector is computed at each time step, combining the current input and the previous hidden state to carry information...

SITE: LEE & TROTT

New Lee & Trott

Our career path is how we spend our time, how we support our lifestyles, how we make our impact, and even sometimes how we define our identity. Let’s make sure we’re on the right track.

Articles about Machine Learning

How does AI work? Each runs off a complex algorithm that tells it what to do and how to learn. Learn how these algorithms work.

Artificial Intelligence (AI) has spread across industries (e.g., business, science, art, education) to enhance user experience, improve work efficienc…

Leading school systems are integrating AI literacy and tools like chatbots and teacher assistants to adapt to how generative AI is reshaping work and learning.

Explore the importance of AI literacy in our AI-driven world. Understand its components, its role in education and business, and how to develop it within organizations.

Interactive series of learning modules for NEIU students, faculty, and staff who want to develop basic skills in AI use

To unpack this, we should first maybe unpack crticial, AI, and literacy. By critical, I mean this in multiple senses of the word. One is critical as in critical thinking, as in skepticism and quest…

Full archive of all the posts from AI + Education = Simplified.