The Transformer architecture underpins modern large language models powering state-of-the-art text generation and AI applications. However, its complexity makes it difficult for non-experts to learn. Existing resources often lack interactivity, rely on static descriptions of simplified architectures, or fail to reflect models’ behavior with real data. To address this gap, we introduce Transformer Explainer, an interactive visualization tool for non-experts to learn Transformers. The tool integrates an overview illustrating the Transformer’s data flow with on-demand explanations that gradually reveal mathematical details. Smooth transitions across abstraction levels highlight the interplay between high-level structures and low-level operations. Running a live GPT-2 instance directly in the browser, Transformer Explainer empowers learners to experiment with custom input and hyperparameters without setup, observing next-token predictions in real time. A 90-participant user study showed that our tool offered significant advantages in improving user understanding and engagement. Transformer Explainer has attracted over 490,000 users.

Transformers have revolutionized machine learning, yet their inner workings remain opaque to many. We present TRANSFORMER EXPLAINER, an interactive visualization tool designed for non-experts to learn about Transformers through the GPT-2 model. Our tool helps users understand complex Transformer concepts by integrating a model overview and smooth transitions across abstraction levels of math operations and model structures. It runs a live GPT-2 model locally in the user’s browser, empowering users to experiment with their own input and observe in real-time how the internal components and parameters of the Transformer work together to predict the next tokens. 125,000 users have used our open-source tool at https://poloclub.github.io/ transformer-explainer/.

Do the rich representations of multi-modal diffusion transformers (DiTs) exhibit unique properties that enhance their interpretability? We introduce ConceptAttention, a novel method that leverages the expressive power of DiT attention layers to generate high-quality saliency maps that precisely locate textual concepts within images. Without requiring additional training, ConceptAttention repurposes the parameters of DiT attention layers to produce highly contextualized *concept embeddings*, contributing the major discovery that performing linear projections in the output space of DiT attention layers yields significantly sharper saliency maps compared to commonly used cross-attention maps. ConceptAttention even achieves state-of-the-art performance on zero-shot image segmentation benchmarks, outperforming 15 other zero-shot interpretability methods on the ImageNet-Segmentation dataset. ConceptAttention works for popular image models and even seamlessly generalizes to video generation. Our work contributes the first evidence that the representations of multi-modal DiTs are highly transferable to vision tasks like segmentation.

We propose a novel energy function for Dense Associative Memory (DenseAM) networks, the log-sum-ReLU (LSR), inspired by optimal kernel density estimation. Unlike the common log-sum-exponential (LSE) function, LSR is based on the Epanechnikov kernel and enables exact memory retrieval with exponential capacity without requiring exponential separation functions. Uniquely, it introduces abundant additional emergent local minima while preserving perfect pattern recovery --- a characteristic previously unseen in DenseAM literature. Empirical results show that LSR energy has significantly more local minima (memories) that have comparable log-likelihood to LSE-based models. Analysis of LSR's emergent memories on image datasets reveals a degree of creativity and novelty, hinting at this method's potential for both large-scale memory storage and generative tasks.

Dense Associative Memories are high storage capacity variants of the Hopfield networks that are capable of storing a large number of memory patterns in the weights of the network of a given size. Their common formulations typically require storing each pattern in a separate set of synaptic weights, which leads to the increase of the number of synaptic weights when new patterns are introduced. In this work we propose an alternative formulation of this class of models using random features, commonly used in kernel methods. In this formulation the number of network's parameters remains fixed. At the same time, new memories can be added to the network by modifying existing weights. We show that this novel network closely approximates the energy function and dynamics of conventional Dense Associative Memories and shares their desirable computational properties.

Diffusion-based generative models’ impressive ability to create convincing images has garnered global attention. However, their complex structures and operations often pose challenges for non-experts to grasp. We present Diffusion Explainer, the first interactive visualization tool that explains how Stable Diffusion transforms text prompts into images. Diffusion Explainer tightly integrates a visual overview of Stable Diffusion’s complex structure with explanations of the underlying operations. By comparing image generation of prompt variants, users can discover the impact of keyword changes on image generation. A 56-participant user study demonstrates that Diffusion Explainer offers substantial learning benefits to non-experts. Our tool has been used by over 10,300 users from 124 countries at https://poloclub.github.io/diffusion-explainer/.

Zijie J. Wang, Evan Montoya, David Munechika, Haoyang Yang, Benjamin Hoover, Duen Horng Chau. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Mitigating algorithmic bias is a critical task in the development and deployment of machine learning models. While several toolkits exist to aid machine learning practitioners in addressing fairness issues, little is known about the strategies practitioners employ to evaluate model fairness and what factors influence their assessment, particularly in the context of text classification. Two common approaches of evaluating the fairness of a model are group fairness and individual fairness. We run a study with Machine Learning practitioners (n=24) to understand the strategies used to evaluate models. Metrics presented to practitioners (group vs. individual fairness) impact which models they consider fair. Participants focused on risks associated with underpredicting / overpredicting and model sensitivity relative to identity token manipulations. We discover fairness assessment strategies involving personal experiences or how users form groups of identity tokens to test model fairness. We provide recommendations for interactive tools for evaluating fairness in text classification.

We present ConceptEvo, a unified interpretation framework for deep neural networks (DNNs) that reveals the inception and evolution of learned concepts during training. Our work addresses a critical gap in DNN interpretation research, as existing methods primarily focus on post-training interpretation. ConceptEvo introduces two novel technical contributions: (1) an algorithm that generates a unified semantic space, enabling side-by-side comparison of different models during training, and (2) an algorithm that discovers and quantifies important concept evolutions for class predictions. Through a large-scale human evaluation and quantitative experiments, we demonstrate that ConceptEvo successfully identifies concept evolutions across different models, which are not only comprehensible to humans but also crucial for class predictions. ConceptEvo is applicable to both modern DNN architectures, such as ConvNeXt, and classic DNNs, such as VGGs and InceptionV3.

Our work combines aspects of three promising paradigms in machine learning, namely, attention mechanism, energy-based models, and associative memory. Attention is the power-house driving modern deep learning successes, but it lacks clear theoretical foundations. Energy-based models allow a principled approach to discriminative and generative tasks, but the design of the energy functional is not straightforward. At the same time, Dense Associative Memory models or Modern Hopfield Networks have a well-established theoretical foundation, and allow an intuitive design of the energy function. We propose a novel architecture, called the Energy Transformer (or ET for short), that uses a sequence of attention layers that are purposely designed to minimize a specifically engineered energy function, which is responsible for representing the relationships between the tokens. In this work, we introduce the theoretical foundations of ET, explore its empirical capabilities using the image completion task, and obtain strong quantitative results on the graph anomaly detection and graph classification tasks.

State-of-the-art neural language models can now be used to solve ad-hoc language tasks through zero-shot prompting without the need for supervised training. This approach has gained popularity in recent years, and researchers have demonstrated prompts that achieve strong accuracy on specific NLP tasks. However, finding a prompt for new tasks requires experimentation. Different prompt templates with different wording choices lead to significant accuracy differences. PromptIDE allows users to experiment with prompt variations, visualize prompt performance, and iteratively optimize prompts. We developed a workflow that allows users to first focus on model feedback using small data before moving on to a large data regime that allows empirical grounding of promising prompts using quantitative measures of the task. The tool then allows easy deployment of the newly created ad-hoc models. We demonstrate the utility of PromptIDE (demo: http://prompt.vizhub.ai) and our workflow using several real-world use cases.

Saliency methods — techniques to identify the importance of input features on a model’s output — are a common step in understanding neural network behavior. However, interpreting saliency requires tedious manual inspection to identify and aggregate patterns in model behavior, resulting in ad hoc or cherry-picked analysis. To address these concerns, we present Shared Interest: metrics for comparing model reasoning (via saliency) to human reasoning (via ground truth annotations). By providing quantitative descriptors, Shared Interest enables ranking, sorting, and aggregating inputs, thereby facilitating large-scale systematic analysis of model behavior. We use Shared Interest to identify eight recurring patterns in model behavior, such as cases where contextual features or a subset of ground truth features are most important to the model. Working with representative real-world users, we show how Shared Interest can be used to decide if a model is trustworthy, uncover issues missed in manual analyses, and enable interactive probing.

AI Security researchers have identified a new way crowdsourced data can be intentionally compromised. Backdoor attacks are a process through which an adversary creates a vulnerability in a machine learning model by ?poisoning?’ the training set by selectively mislabelling images containing a backdoor object. The model continues to perform well on standard testing data but misclassifies on the inputs that contain the backdoor chosen by the adversary. In this paper, we present the design and development of the Backdoor Game, the first game in which users can interact with different poisoned classifiers and upload their own images containing backdoor objects in an engaging way. We conduct semi-structured interviews with eight different participants who interacted with a first version of the Backdoor Game and deploy the game to Mechanical Turk users (N=68) to demonstrate how users interacted with the backdoor objects. We present results including novel types of interactions that emerged as a result of game play and design recommendations for the improvement of the system. The combined design, development and deployment of our system can help AI Security researchers to study this emerging concept, from determining the effectiveness of different backdoor objects to help compiling a collection of diverse and unique backdoor objects from the public, increasing the safety of future AI systems.

The spread of infectious diseases such as COVID-19 presents many challenges to healthcare systems and infrastructures across the world, exacerbating inequalities and leaving the world's most vulnerable populations most affected. Given their density and available infrastructure, refugee and internally displaced person (IDP) settlements can be particularly susceptible to disease spread. In this paper we present an agent-based modeling approach to simulating the spread of disease in refugee and IDP settlements under various non-pharmaceutical intervention strategies. The model, based on the June open-source framework, is informed by data on geography, demographics, comorbidities, physical infrastructure and other parameters obtained from real-world observations and previous literature. The development and testing of this approach focuses on the Cox's Bazar refugee settlement in Bangladesh, although our model is designed to be generalizable to other informal settings. Our findings suggest the encouraging self-isolation at home of mild to severe symptomatic patients, as opposed to the isolation of all positive cases in purpose-built isolation and treatment centers, does not increase the risk of secondary infection meaning the centers can be used to provide hospital support to the most intense cases of COVID-19. Secondly we find that mask wearing in all indoor communal areas can be effective at dampening viral spread, even with low mask efficacy and compliance rates. Finally, we model the effects of reopening learning centers in the settlement under various mitigation strategies. For example, a combination of mask wearing in the classroom, halving attendance regularity to enable physical distancing, and better ventilation can almost completely mitigate the increased risk of infection which keeping the learning centers open may cause. These modeling efforts are being incorporated into decision making processes to inform future planning, and further exercises should be carried out in similar geographies to help protect those most vulnerable.

The novel nature of SARS-CoV-2 calls for the development of efficient de novo drug design approaches. In this study, we propose an end-to-end framework, named CogMol (Controlled Generation of Molecules), for designing new drug-like small molecules targeting novel viral proteins with high affinity and off-target selectivity. CogMol combines adaptive pre-training of a molecular SMILES Variational Autoencoder (VAE) and an efficient multi-attribute controlled sampling scheme that uses guidance from attribute predictors trained on latent features. To generate novel and optimal drug-like molecules for unseen viral targets, CogMol leverages a protein-molecule binding affinity predictor that is trained using SMILES VAE embeddings and protein sequence embeddings learned unsupervised from a large corpus. We applied the CogMol framework to three SARS-CoV-2 target proteins: main protease, receptor-binding domain of the spike protein, and non-structural protein 9 replicase. The generated candidates are novel at both the molecular and chemical scaffold levels when compared to the training data. CogMol also includes insilico screening for assessing toxicity of parent molecules and their metabolites with a multi-task toxicity classifier, synthetic feasibility with a chemical retrosynthesis predictor, and target structure binding with docking simulations. Docking reveals favorable binding of generated molecules to the target protein structure, where 87--95\% of high affinity molecules showed docking free energy $<$ -6 kcal/mol. When compared to approved drugs, the majority of designed compounds show low predicted parent molecule and metabolite toxicity and high predicted synthetic feasibility. In summary, CogMol can handle multi-constraint design of synthesizable, low-toxic, drug-like molecules with high target specificity and selectivity, even to novel protein target sequences, and does not need target-dependent fine-tuning of the framework or target structure information.