Abstract:While large language models (LLMs) have shown remarkable capability to generate convincing text across diverse domains, concerns around its potential risks have highlighted the importance of understanding the rationale behind text generation. We present LLM Attributor, a Python library that provides interactive visualizations for training data attribution of an LLM's text generation. Our library offers a new way to quickly attribute an LLM's text generation to training data points to inspect model behaviors, enhance its trustworthiness, and compare model-generated text with user-provided text. We describe the visual and interactive design of our tool and highlight usage scenarios for LLaMA2 models fine-tuned with two different datasets: online articles about recent disasters and finance-related question-answer pairs. Thanks to LLM Attributor's broad support for computational notebooks, users can easily integrate it into their workflow to interactively visualize attributions of their models. For easier access and extensibility, we open-source LLM Attributor at https://github.com/poloclub/ LLM-Attribution. The video demo is available at https://youtu.be/mIG2MDQKQxM.
Abstract:Our research aims to unify existing works' diverging opinions on how architectural components affect the adversarial robustness of CNNs. To accomplish our goal, we synthesize a suite of three generalizable robust architectural design principles: (a) optimal range for depth and width configurations, (b) preferring convolutional over patchify stem stage, and (c) robust residual block design through adopting squeeze and excitation blocks and non-parametric smooth activation functions. Through extensive experiments across a wide spectrum of dataset scales, adversarial training methods, model parameters, and network design spaces, our principles consistently and markedly improve AutoAttack accuracy: 1-3 percentage points (pp) on CIFAR-10 and CIFAR-100, and 4-9 pp on ImageNet. The code is publicly available at https://github.com/poloclub/robust-principles.
Abstract:Large language models (LLMs) have skyrocketed in popularity in recent years due to their ability to generate high-quality text in response to human prompting. However, these models have been shown to have the potential to generate harmful content in response to user prompting (e.g., giving users instructions on how to commit crimes). There has been a focus in the literature on mitigating these risks, through methods like aligning models with human values through reinforcement learning. However, it has been shown that even aligned language models are susceptible to adversarial attacks that bypass their restrictions on generating harmful text. We propose a simple approach to defending against these attacks by having a large language model filter its own responses. Our current results show that even if a model is not fine-tuned to be aligned with human values, it is possible to stop it from presenting harmful content to users by validating the content using a language model.