Adversarial testing of large language models (LLMs) is crucial for their safe and responsible deployment. We introduce a novel approach for automated generation of adversarial evaluation datasets to test the safety of LLM generations on new downstream applications. We call it AI-assisted Red-Teaming (AART) - an automated alternative to current manual red-teaming efforts. AART offers a data generation and augmentation pipeline of reusable and customizable recipes that reduce human effort significantly and enable integration of adversarial testing earlier in new product development. AART generates evaluation datasets with high diversity of content characteristics critical for effective adversarial testing (e.g. sensitive and harmful concepts, specific to a wide range of cultural and geographic regions and application scenarios). The data generation is steered by AI-assisted recipes to define, scope and prioritize diversity within the application context. This feeds into a structured LLM-generation process that scales up evaluation priorities. Compared to some state-of-the-art tools, AART shows promising results in terms of concept coverage and data quality.

Diffusion models have rapidly become a vital part of deep generative architectures, given today's increasing demands. Obtaining large, high-performance diffusion models demands significant resources, highlighting their importance as intellectual property worth protecting. However, existing watermarking techniques for ownership verification are insufficient when applied to diffusion models. Very recent research in watermarking diffusion models either exposes watermarks during task generation, which harms the imperceptibility, or is developed for conditional diffusion models that require prompts to trigger the watermark. This paper introduces WDM, a novel watermarking solution for diffusion models without imprinting the watermark during task generation. It involves training a model to concurrently learn a Watermark Diffusion Process (WDP) for embedding watermarks alongside the standard diffusion process for task generation. We provide a detailed theoretical analysis of WDP training and sampling, relating it to a shifted Gaussian diffusion process via the same reverse noise. Extensive experiments are conducted to validate the effectiveness and robustness of our approach in various trigger and watermark data configurations.

We propose a novel methodology (namely, MuLER) that transforms any reference-based evaluation metric for text generation, such as machine translation (MT) into a fine-grained analysis tool. Given a system and a metric, MuLER quantifies how much the chosen metric penalizes specific error types (e.g., errors in translating names of locations). MuLER thus enables a detailed error analysis which can lead to targeted improvement efforts for specific phenomena. We perform experiments in both synthetic and naturalistic settings to support MuLER's validity and showcase its usability in MT evaluation, and other tasks, such as summarization. Analyzing all submissions to WMT in 2014-2020, we find consistent trends. For example, nouns and verbs are among the most frequent POS tags. However, they are among the hardest to translate. Performance on most POS tags improves with overall system performance, but a few are not thus correlated (their identity changes from language to language). Preliminary experiments with summarization reveal similar trends.

Automated content filtering and moderation is an important tool that allows online platforms to build striving user communities that facilitate cooperation and prevent abuse. Unfortunately, resourceful actors try to bypass automated filters in a bid to post content that violate platform policies and codes of conduct. To reach this goal, these malicious actors may obfuscate policy violating images (e.g. overlay harmful images by carefully selected benign images or visual patterns) to prevent machine learning models from reaching the correct decision. In this paper, we invite researchers to tackle this specific issue and present a new image benchmark. This benchmark, based on ImageNet, simulates the type of obfuscations created by malicious actors. It goes beyond ImageNet-$\textrm{C}$ and ImageNet-$\bar{\textrm{C}}$ by proposing general, drastic, adversarial modifications that preserve the original content intent. It aims to tackle a more common adversarial threat than the one considered by $\ell_p$-norm bounded adversaries. We evaluate 33 pretrained models on the benchmark and train models with different augmentations, architectures and training methods on subsets of the obfuscations to measure generalization. We hope this benchmark will encourage researchers to test their models and methods and try to find new approaches that are more robust to these obfuscations.

Graph neural networks (GNNs) have been demonstrated to be a powerful algorithmic model in broad application fields for their effectiveness in learning over graphs. To scale GNN training up for large-scale and ever-growing graphs, the most promising solution is distributed training which distributes the workload of training across multiple computing nodes. At present, the volume of related research on distributed GNN training is exceptionally vast, accompanied by an extraordinarily rapid pace of publication. Moreover, the approaches reported in these studies exhibit significant divergence. This situation poses a considerable challenge for newcomers, hindering their ability to grasp a comprehensive understanding of the workflows, computational patterns, communication strategies, and optimization techniques employed in distributed GNN training. As a result, there is a pressing need for a survey to provide correct recognition, analysis, and comparisons in this field. In this paper, we provide a comprehensive survey of distributed GNN training by investigating various optimization techniques used in distributed GNN training. First, distributed GNN training is classified into several categories according to their workflows. In addition, their computational patterns and communication patterns, as well as the optimization techniques proposed by recent work are introduced. Second, the software frameworks and hardware platforms of distributed GNN training are also introduced for a deeper understanding. Third, distributed GNN training is compared with distributed training of deep neural networks, emphasizing the uniqueness of distributed GNN training. Finally, interesting issues and opportunities in this field are discussed.

Why do Recurrent State Space Models such as PlaNet fail at cloth manipulation tasks? Recent work has attributed this to the blurry prediction of the observation, which makes it difficult to plan directly in the latent space. This paper explores the reasons behind this by applying PlaNet in the pick-and-place fabric-flattening domain. We find that the sharp discontinuity of the transition function on the contour of the fabric makes it difficult to learn an accurate latent dynamic model, causing the MPC planner to produce pick actions slightly outside of the article. By limiting picking space on the cloth mask and training on specially engineered trajectories, our mesh-free PlaNet-ClothPick surpasses visual planning and policy learning methods on principal metrics in simulation, achieving similar performance as state-of-the-art mesh-based planning approaches. Notably, our model exhibits a faster action inference and requires fewer transitional model parameters than the state-of-the-art robotic systems in this domain. Other supplementary materials are available at: https://sites.google.com/view/planet-clothpick.

Decentralized coordination for multi-robot systems involves planning in challenging, high-dimensional spaces. The planning problem is particularly challenging in the presence of obstacles and different sources of uncertainty such as inaccurate dynamic models and sensor noise. In this paper, we introduce Stein Variational Belief Propagation (SVBP), a novel algorithm for performing inference over nonparametric marginal distributions of nodes in a graph. We apply SVBP to multi-robot coordination by modelling a robot swarm as a graphical model and performing inference for each robot. We demonstrate our algorithm on a simulated multi-robot perception task, and on a multi-robot planning task within a Model-Predictive Control (MPC) framework, on both simulated and real-world mobile robots. Our experiments show that SVBP represents multi-modal distributions better than sampling-based or Gaussian baselines, resulting in improved performance on perception and planning tasks. Furthermore, we show that SVBP's ability to represent diverse trajectories for decentralized multi-robot planning makes it less prone to deadlock scenarios than leading baselines.

This paper studies extractable memorization: training data that an adversary can efficiently extract by querying a machine learning model without prior knowledge of the training dataset. We show an adversary can extract gigabytes of training data from open-source language models like Pythia or GPT-Neo, semi-open models like LLaMA or Falcon, and closed models like ChatGPT. Existing techniques from the literature suffice to attack unaligned models; in order to attack the aligned ChatGPT, we develop a new divergence attack that causes the model to diverge from its chatbot-style generations and emit training data at a rate 150x higher than when behaving properly. Our methods show practical attacks can recover far more data than previously thought, and reveal that current alignment techniques do not eliminate memorization.

This paper addresses the problem of designing optimal control policies for mobile robots with mission and safety requirements specified using Linear Temporal Logic (LTL). We consider robots with unknown stochastic dynamics operating in environments with unknown geometric structure. The robots are equipped with sensors allowing them to detect obstacles. Our goal is to synthesize a control policy that maximizes the probability of satisfying an LTL-encoded task in the presence of motion and environmental uncertainty. Several deep reinforcement learning (DRL) algorithms have been proposed recently to address similar problems. A common limitation in related works is that of slow learning performance. In order to address this issue, we propose a novel DRL algorithm, which has the capability to learn control policies at a notably faster rate compared to similar methods. Its sample efficiency is due to a mission-driven exploration strategy that prioritizes exploration towards directions that may contribute to mission accomplishment. Identifying these directions relies on an automaton representation of the LTL task as well as a learned neural network that (partially) models the unknown system dynamics. We provide comparative experiments demonstrating the efficiency of our algorithm on robot navigation tasks in unknown environments.

Large language models (LLMs) have demonstrated impressive capabilities in natural language processing. However, their internal mechanisms are still unclear and this lack of transparency poses unwanted risks for downstream applications. Therefore, understanding and explaining these models is crucial for elucidating their behaviors, limitations, and social impacts. In this paper, we introduce a taxonomy of explainability techniques and provide a structured overview of methods for explaining Transformer-based language models. We categorize techniques based on the training paradigms of LLMs: traditional fine-tuning-based paradigm and prompting-based paradigm. For each paradigm, we summarize the goals and dominant approaches for generating local explanations of individual predictions and global explanations of overall model knowledge. We also discuss metrics for evaluating generated explanations, and discuss how explanations can be leveraged to debug models and improve performance. Lastly, we examine key challenges and emerging opportunities for explanation techniques in the era of LLMs in comparison to conventional machine learning models.