Previous studies have typically assumed that large language models are unable to accurately perform arithmetic operations, particularly multiplication of >8 digits, and operations involving decimals and fractions, without the use of calculator tools. This paper aims to challenge this misconception. With sufficient training data, a 2 billion-parameter language model can accurately perform multi-digit arithmetic operations with almost 100% accuracy without data leakage, significantly surpassing GPT-4 (whose multi-digit multiplication accuracy is only 4.3%). We also demonstrate that our MathGLM, fine-tuned from GLM-10B on a dataset with additional multi-step arithmetic operations and math problems described in text, achieves similar performance to GPT-4 on a 5,000-samples Chinese math problem test set. Our code and data are public at https://github.com/THUDM/MathGLM.

This paper deals with a network of computing agents aiming to solve an online optimization problem in a distributed fashion, i.e., by means of local computation and communication, without any central coordinator. We propose the gradient tracking with adaptive momentum estimation (GTAdam) distributed algorithm, which combines a gradient tracking mechanism with first and second order momentum estimates of the gradient. The algorithm is analyzed in the online setting for strongly convex cost functions with Lipschitz continuous gradients. We provide an upper bound for the dynamic regret given by a term related to the initial conditions and another term related to the temporal variations of the objective functions. Moreover, a linear convergence rate is guaranteed in the static setup. The algorithm is tested on a time-varying classification problem, on a (moving) target localization problem, and in a stochastic optimization setup from image classification. In these numerical experiments from multi-agent learning, GTAdam outperforms state-of-the-art distributed optimization methods.

Training energy-based models (EBMs) with maximum likelihood estimation on high-dimensional data can be both challenging and time-consuming. As a result, there a noticeable gap in sample quality between EBMs and other generative frameworks like GANs and diffusion models. To close this gap, inspired by the recent efforts of learning EBMs by maximimizing diffusion recovery likelihood (DRL), we propose cooperative diffusion recovery likelihood (CDRL), an effective approach to tractably learn and sample from a series of EBMs defined on increasingly noisy versons of a dataset, paired with an initializer model for each EBM. At each noise level, the initializer model learns to amortize the sampling process of the EBM, and the two models are jointly estimated within a cooperative training framework. Samples from the initializer serve as starting points that are refined by a few sampling steps from the EBM. With the refined samples, the EBM is optimized by maximizing recovery likelihood, while the initializer is optimized by learning from the difference between the refined samples and the initial samples. We develop a new noise schedule and a variance reduction technique to further improve the sample quality. Combining these advances, we significantly boost the FID scores compared to existing EBM methods on CIFAR-10 and ImageNet 32x32, with a 2x speedup over DRL. In addition, we extend our method to compositional generation and image inpainting tasks, and showcase the compatibility of CDRL with classifier-free guidance for conditional generation, achieving similar trade-offs between sample quality and sample diversity as in diffusion models.

Sharifa Alghowinem, a research scientist in the Media Lab's Personal Robots Group, poses with Jibo, a friendly robot companion developed by Professor Cynthia Breazeal. "As a child, I wished for a robot that would explain others' emotions to me" says Sharifa Alghowinem, a research scientist in the Media Lab's Personal Robots Group (PRG). Growing up in Saudi Arabia, Alghowinem says she dreamed of coming to MIT one day to develop Arabic-based technologies, and of creating a robot that could help herself and others navigate a complex world. In her early life, Alghowinem faced difficulties with understanding social cues and never scored well on standardized tests, but her dreams carried her through. She earned an undergraduate degree in computing before leaving home to pursue graduate education in Australia.

In the context of kernel machines, polynomial and Fourier features are commonly used to provide a nonlinear extension to linear models by mapping the data to a higher-dimensional space. Unless one considers the dual formulation of the learning problem, which renders exact large-scale learning unfeasible, the exponential increase of model parameters in the dimensionality of the data caused by their tensor-product structure prohibits to tackle high-dimensional problems. One of the possible approaches to circumvent this exponential scaling is to exploit the tensor structure present in the features by constraining the model weights to be an underparametrized tensor network. In this paper we quantize, i.e. further tensorize, polynomial and Fourier features. Based on this feature quantization we propose to quantize the associated model weights, yielding quantized models. We show that, for the same number of model parameters, the resulting quantized models have a higher bound on the VC-dimension as opposed to their non-quantized counterparts, at no additional computational cost while learning from identical features. We verify experimentally how this additional tensorization regularizes the learning problem by prioritizing the most salient features in the data and how it provides models with increased generalization capabilities. We finally benchmark our approach on large regression task, achieving state-of-the-art results on a laptop computer.

This article summarizes a systematic review of the electroencephalography (EEG)-based cognitive workload (CWL) estimation. The focus of the article is twofold: identify the disparate experimental paradigms used for reliably eliciting discreet and quantifiable levels of cognitive load and the specific nature and representational structure of the commonly used input formulations in deep neural networks (DNNs) used for signal classification. The analysis revealed a number of studies using EEG signals in its native representation of a two-dimensional matrix for offline classification of CWL. However, only a few studies adopted an online or pseudo-online classification strategy for real-time CWL estimation. Further, only a couple of interpretable DNNs and a single generative model were employed for cognitive load detection till date during this review. More often than not, researchers were using DNNs as black-box type models. In conclusion, DNNs prove to be valuable tools for classifying EEG signals, primarily due to the substantial modeling power provided by the depth of their network architecture. It is further suggested that interpretable and explainable DNN models must be employed for cognitive workload estimation since existing methods are limited in the face of the non-stationary nature of the signal.

This paper presents a pipeline with minimal human influence for scraping and detecting bias on college newspaper archives. This paper introduces a framework for scraping complex archive sites that automated tools fail to grab data from, and subsequently generates a dataset of 14 student papers with 23,154 entries. This data can also then be queried by keyword to calculate bias by comparing the sentiment of a large language model summary to the original article. The advantages of this approach are that it is less comparative than reconstruction bias and requires less labelled data than generating keyword sentiment. Results are calculated on politically charged words as well as control words to show how conclusions can be drawn. The complete method facilitates the extraction of nuanced insights with minimal assumptions and categorizations, paving the way for a more objective understanding of bias within student newspaper sources.

Transformers have become the dominant model in deep learning, but the reason for their superior performance is poorly understood. Here, we hypothesize that the strong performance of Transformers stems from an architectural bias towards mesa-optimization, a learned process running within the forward pass of a model consisting of the following two steps: (i) the construction of an internal learning objective, and (ii) its corresponding solution found through optimization. To test this hypothesis, we reverse-engineer a series of autoregressive Transformers trained on simple sequence modeling tasks, uncovering underlying gradient-based mesa-optimization algorithms driving the generation of predictions. Moreover, we show that the learned forward-pass optimization algorithm can be immediately repurposed to solve supervised few-shot tasks, suggesting that mesa-optimization might underlie the in-context learning capabilities of large language models. Finally, we propose a novel self-attention layer, the mesa-layer, that explicitly and efficiently solves optimization problems specified in context. We find that this layer can lead to improved performance in synthetic and preliminary language modeling experiments, adding weight to our hypothesis that mesa-optimization is an important operation hidden within the weights of trained Transformers.

Semi-supervised learning (SSL) has become a crucial approach in deep learning as a way to address the challenge of limited labeled data. The success of deep neural networks heavily relies on the availability of large-scale high-quality labeled data. However, the process of data labeling is time-consuming and unscalable, leading to shortages in labeled data. SSL aims to tackle this problem by leveraging additional unlabeled data in the training process. One of the popular SSL algorithms, FixMatch, trains identical weight-sharing teacher and student networks simultaneously using a siamese neural network (SNN). However, it is prone to performance degradation when the pseudo labels are heavily noisy in the early training stage. We present KD-FixMatch, a novel SSL algorithm that addresses the limitations of FixMatch by incorporating knowledge distillation. The algorithm utilizes a combination of sequential and simultaneous training of SNNs to enhance performance and reduce performance degradation. Firstly, an outer SNN is trained using labeled and unlabeled data. After that, the network of the well-trained outer SNN generates pseudo labels for the unlabeled data, from which a subset of unlabeled data with trusted pseudo labels is then carefully created through high-confidence sampling and deep embedding clustering. Finally, an inner SNN is trained with the labeled data, the unlabeled data, and the subset of unlabeled data with trusted pseudo labels. Experiments on four public data sets demonstrate that KD-FixMatch outperforms FixMatch in all cases. Our results indicate that KD-FixMatch has a better training starting point that leads to improved model performance compared to FixMatch.

Online machine learning (ML) is often used in self-adaptive systems to strengthen the adaptation mechanism and improve the system utility. Despite such benefits, applying online ML for self-adaptation can be challenging, and not many papers report its limitations. Recently, we experimented with applying online ML for self-adaptation of a smart farming scenario and we had faced several unexpected difficulties -- traps -- that, to our knowledge, are not discussed enough in the community. In this paper, we report our experience with these traps. Specifically, we discuss several traps that relate to the specification and online training of the ML-based estimators, their impact on self-adaptation, and the approach used to evaluate the estimators. Our overview of these traps provides a list of lessons learned, which can serve as guidance for other researchers and practitioners when applying online ML for self-adaptation.