Dataset distillation aims to learn a small synthetic dataset that preserves most of the information from the original dataset. Dataset distillation can be formulated as a bi-level meta-learning problem where the outer loop optimizes the meta-dataset and the inner loop trains a model on the distilled data. Meta-gradient computation is one of the key challenges in this formulation, as differentiating through the inner loop learning procedure introduces significant computation and memory costs. In this paper, we address these challenges using neural Feature Regression with Pooling (FRePo), achieving the state-of-the-art performance with an order of magnitude less memory requirement and two orders of magnitude faster training than previous methods. The proposed algorithm is analogous to truncated backpropagation through time with a pool of models to alleviate various types of overfitting in dataset distillation. FRePo significantly outperforms the previous methods on CIFAR100, Tiny ImageNet, and ImageNet-1K. Furthermore, we show that high-quality distilled data can greatly improve various downstream applications, such as continual learning and membership inference defense. Please check out our webpage at https://sites.google.com/view/frepo.

Argument mining tasks require an informed range of low to high complexity linguistic phenomena and commonsense knowledge. Previous work has shown that pre-trained language models are highly effective at encoding syntactic and semantic linguistic phenomena when applied with transfer learning techniques and built on different pre-training objectives. It remains an issue of how much the existing pre-trained language models encompass the complexity of argument mining tasks. We rely on experimentation to shed light on how language models obtained from different lexical semantic families leverage the performance of the identification of argumentative discourse units task. Experimental results show that transfer learning techniques are beneficial to the task and that current methods may be insufficient to leverage commonsense knowledge from different lexical semantic families.

Despite their strong performance on many tasks, pre-trained language models have been shown to struggle on out-of-distribution compositional generalization. Meanwhile, recent work has shown considerable improvements on many NLP tasks from model scaling. Can scaling up model size also improve compositional generalization in semantic parsing? We evaluate encoder-decoder models up to 11B parameters and decoder-only models up to 540B parameters, and compare model scaling curves for three different methods for applying a pre-trained language model to a new task: fine-tuning all parameters, prompt tuning, and in-context learning. We observe that fine-tuning generally has flat or negative scaling curves on out-of-distribution compositional generalization in semantic parsing evaluations. In-context learning has positive scaling curves, but is generally outperformed by much smaller fine-tuned models. Prompt-tuning can outperform fine-tuning, suggesting further potential improvements from scaling as it exhibits a more positive scaling curve. Additionally, we identify several error trends that vary with model scale. For example, larger models are generally better at modeling the syntax of the output space, but are also more prone to certain types of overfitting. Overall, our study highlights limitations of current techniques for effectively leveraging model scale for compositional generalization, while our analysis also suggests promising directions for future work.

Adaptive curricula in reinforcement learning (RL) have proven effective for producing policies robust to discrepancies between the train and test environment. Recently, the Unsupervised Environment Design (UED) framework generalized RL curricula to generating sequences of entire environments, leading to new methods with robust minimax regret properties. Problematically, in partially-observable or stochastic settings, optimal policies may depend on the ground-truth distribution over aleatoric parameters of the environment in the intended deployment setting, while curriculum learning necessarily shifts the training distribution. We formalize this phenomenon as curriculum-induced covariate shift (CICS), and describe how its occurrence in aleatoric parameters can lead to suboptimal policies. Directly sampling these parameters from the ground-truth distribution avoids the issue, but thwarts curriculum learning. We propose SAMPLR, a minimax regret UED method that optimizes the ground-truth utility function, even when the underlying training data is biased due to CICS. We prove, and validate on challenging domains, that our approach preserves optimality under the ground-truth distribution, while promoting robustness across the full range of environment settings.

QA models with lifelong learning (LL) abilities are important for practical QA applications, and architecture-based LL methods are reported to be an effective implementation for these models. However, it is non-trivial to extend previous approaches to QA tasks since they either require access to task identities in the testing phase or do not explicitly model samples from unseen tasks. In this paper, we propose Diana: a dynamic architecture-based lifelong QA model that tries to learn a sequence of QA tasks with a prompt enhanced language model. Four types of hierarchically organized prompts are used in Diana to capture QA knowledge from different granularities. Specifically, we dedicate task-level prompts to capture task-specific knowledge to retain high LL performances and maintain instance-level prompts to learn knowledge shared across different input samples to improve the model's generalization performance. Moreover, we dedicate separate prompts to explicitly model unseen tasks and introduce a set of prompt key vectors to facilitate knowledge sharing between tasks. Extensive experiments demonstrate that Diana outperforms state-of-the-art lifelong QA models, especially in handling unseen tasks.

Over the last years, word and sentence embeddings have established as text preprocessing for all kinds of NLP tasks and improved performances in these tasks significantly. Unfortunately, it has also been shown that these embeddings inherit various kinds of biases from the training data and thereby pass on biases present in society to NLP solutions. Many papers attempted to quantify bias in word or sentence embeddings to evaluate debiasing methods or compare different embedding models, often with cosine-based scores. However, some works have raised doubts about these scores showing that even though they report low biases, biases persist and can be shown with other tests. In fact, there is a great variety of bias scores or tests proposed in the literature without any consensus on the optimal solutions. We lack works that study the behavior of bias scores and elaborate their advantages and disadvantages. In this work, we will explore different cosine-based bias scores. We provide a bias definition based on the ideas from the literature and derive novel requirements for bias scores. Furthermore, we thoroughly investigate the existing cosine-based scores and their limitations in order to show why these scores fail to report biases in some situations. Finally, we propose a new bias score, SAME, to address the shortcomings of existing bias scores and show empirically that SAME is better suited to quantify biases in word embeddings.

This article has been excerpted from The Equality Machine: Harnessing Digital Technology for a Brighter, More Inclusive Future by Orly Lobel. For years, Japan has been the indisputable leader in robotics. If Tanzania's Olduvai Gorge is the cradle of humanity, Japan is the cradle of the humanoids, developing the first humanoid robot in the 1970s and many iterations since. Japanese roboticists pioneered the notion that artificial intelligence should be embodied. While the West focused more on algorithms in the abstract, Japanese institutions believed that AI innovation should be developed alongside--or rather, within--a physical artificial body. Japanese roboticists have been leading the way in realizing the aspiration to create robots that offer companionship to humans for decades.

Abstract: Quantum computers have the potential to speed up certain computational tasks. A possibility this opens up within the field of machine learning is the use of quantum techniques that may be inefficient to simulate classically but could provide superior performance in some tasks. Machine learning algorithms are ubiquitous in particle physics and as advances are made in quantum machine learning technology there may be a similar adoption of these quantum techniques. In this work a quantum support vector machine (QSVM) is implemented for signal-background classification. We investigate the effect of different quantum encoding circuits, the process that transforms classical data into a quantum state, on the final classification performance.

Staring closely at the screen, Jordan Wright deftly picks out a barely distinguishable shape with his mouse, bringing to life a stark blue outline from a blur of overexposed features. It is a process similar to the automated tests that teach computers to distinguish humans from machines, by asking someone to identify traffic lights or stop signs in a picture known as a Captcha. Only in Wright's case, the shape turns out to be of a Tupolev Tu-160, a supersonic strategic heavy bomber, parked on a Russian base. The outline -- one of hundreds a day he picks out from satellite images -- is training an algorithm so that a US intelligence agency can locate and identify Moscow's firepower in an automated flash. It has become a run-of-the-mill task for the 25-year-old, who describes himself as on the autism spectrum. Starting in the spring, Wright began working at Enabled Intelligence Inc, a Virginia-based start-up that works largely for US intelligence and other federal agencies.