A.1 Creation of the Multimodal Web Document Dataset A.1.1 Collecting of a Large Number of HTMLFiles Our data collection process begins by considering the 25 most recent Common Crawl6 dumps available at the time of dataset creation. It contains webpages spanning from February 2020 to January/February 2023. We use a modified version of readability-lxml7 to extract the main text from the pages, discarding any pages that contain text of excessively high perplexity. This process yields a total of 41.2 billion documents. Selection of English content To identify non-English content, we apply the FastText classifier (Joulin et al., 2017) to the extracted text, e ectively filtering out 63.6% of the documents. Early text deduplication Often, a set of URLs is crawled repeatedly across di erent Common Crawl snapshots. However, the content of these websites may vary as web administrators make changes over time. Hence, at this stage, we refrain from deduplicating documents based on their URLs. Instead, we perform MinHash (Broder, 1997) deduplication with 16 hashes calculated over 5-grams. To further refine the data, we eliminate documents containing substantial proportions of repeated paragraphs and n-grams, employing the methodology described in MassiveText (Rae et al., 2022).

Computer-Use Agents (CUA) are becoming increasingly capable of autonomously operating digital environments through Graphical User Interfaces (GUI). Yet, most GUI remain designed primarily for humans--prioritizing aesthetics and usability--forcing agents to adopt human-oriented behaviors that are unnecessary for efficient task execution. At the same time, rapid advances in coding-oriented language models (Coder) have transformed automatic GUI design. This raises a fundamental question: Can CUA as judges to assist Coder for automatic GUI design? To investigate, we introduce AUI-Gym, a benchmark for Automatic GUI development spanning 52 applications across diverse domains. Using language models, we synthesize 1560 tasks that simulate real-world scenarios. To ensure task reliability, we further develop a verifier that programmatically checks whether each task is executable within its environment. Building on this, we propose a Coder-CUA in Collaboration framework: the Coder acts as Designer, generating and revising websites, while the CUA serves as Judge, evaluating functionality and refining designs. Success is measured not by visual appearance, but by task solvability and CUA navigation success rate. To turn CUA feedback into usable guidance, we design a CUA Dashboard that compresses multi-step navigation histories into concise visual summaries, offering interpretable guidance for iterative redesign. By positioning agents as both designers and judges, our framework shifts interface design toward agent-native efficiency and reliability. Our work takes a step toward shifting agents from passive use toward active participation in digital environments. Our code and dataset are available at https://github.com/showlab/AUI.

WebChecker is a plugin for Epsilon Validation Language (EVL), designed to validate both static and dynamic HTML pages utilizing frameworks like Bootstrap. By employing configurable EVL constraints, WebChecker enforces implicit rules governing HTML and CSS frameworks. The effectiveness of the plugin is demonstrated through its application on Bootstrap, the widely adopted HTML, CSS, and JavaScript framework. WebChecker comes with a set of EVL constraints to assess Bootstrap based web pages. To substantiate our claims, I present an illustrative example featuring two solutions that effectively enforce implicit rules.

The original Chapter 10 lab made use of keras, an R package for deep learning that relies on Python. Getting keras to work on your computer can be a bit of a challenge. Installation instructions are available here. RStudio has recently released a new R package for deep learning, called torch, that does not require a Python installation. Daniel Falbel and Sigrid Keydana, two of the torch developers, translated our keras version of the Chapter 10 lab to torch.

In this project, I try to use a machine learning package, MLSeq from Bioconductor to find out the best model that predicts breast cancer subtype. I use 28 data sets from TCGA to train and test this model using 12 datasets. So I may do many copy and paste, but I will give my understanding and opinions in the R notebook. According to the Vignette, I've input the data and converted the data to be right data frames which are ready to do MLSeq. And the next step is to choose a model, do the Normalization and transformation, and use the normalized data to train model.

When ChatGPT came out at the end of 2022, It caused a huge surprise for millions of people around the world due to its capabilities. I experienced the rise of it firsthand: I have personally seen the reactions of people using it and have had numerous interactions with GPT. For some people, It may even sound surreal to build a game with ChatGPT without writing code, but it is possible now. In case you have never heard it before, ChatGPT is a super powerful AI tool created by OpenAI, a company founded in San Francisco in late 2015 by Sam Altman, Elon Musk, and others. You can take a look here to learn its capabilities briefly.

Transformers Interpret is a model explainability tool designed to work exclusively with the transformers package. In line with the philosophy of the Transformers package Transformers Interpret allows any transformers model to be explained in just two lines. Explainers are available for both text and computer vision models. Visualizations are also available in notebooks and as savable png and html files. Positive attribution numbers indicate a word contributes positively towards the predicted class, while negative numbers indicate a word contributes negatively towards the predicted class.

While not everyone's convinced that AI art is actual art, the generators used to whip them up are likely here to stay. DeviantArt is now getting into the space with a generator of its own called DreamUp, promising "safe and fair" generation for creators. The website says one of artists' main concerns about AI art is that their work may be used to train artificial intelligence models, which means the generator could spit out pieces in their style without their consent. In an attempt to give artists control over their work, DeviantArt is giving them the ability to choose whether or not the tool can use their style for direct inspiration. Further, the website is giving them the power to declare whether or not to allow their work to be used in datasets used to train third-party AI models.

Whenever we debate with other devs about the best programming language, we talk about JavaScript and Python for hours. Both are powerful, flexible languages that are dominating the world today. But a dead end to Python is its inability to run on browsers. JavaScript (JS), with the discovery of Node, runs on almost any platform. It even has modules to build machine learning algorithms.