By providing external information to large language models (LLMs), tool augmentation (including retrieval augmentation) has emerged as a promising solution for addressing the limitations of LLMs' static parametric memory. However, how receptive are LLMs to such external evidence, especially when the evidence conflicts with their parametric memory? We present the first comprehensive and controlled investigation into the behavior of LLMs when encountering knowledge conflicts. We propose a systematic framework to elicit high-quality parametric memory from LLMs and construct the corresponding counter-memory, which enables us to conduct a series of controlled experiments. Our investigation reveals seemingly contradicting behaviors of LLMs. On the one hand, different from prior wisdom, we find that LLMs can be highly receptive to external evidence even when that conflicts with their parametric memory, given that the external evidence is coherent and convincing. On the other hand, LLMs also demonstrate a strong confirmation bias when the external evidence contains some information that is consistent with their parametric memory, despite being presented with conflicting evidence at the same time. These results pose important implications that are worth careful consideration for the further development and deployment of tool- and retrieval-augmented LLMs.

The generation of molecules with desired properties has gained tremendous popularity, revolutionizing the way scientists design molecular structures and providing valuable support for chemical and drug design. However, despite the potential of language models in molecule generation, they face numerous challenges such as the generation of syntactically or chemically flawed molecules, narrow domain focus, and limitations in creating diverse and directionally feasible molecules due to a dearth of annotated data or external molecular databases. To tackle these challenges, we introduce MolGen, a pre-trained molecular language model tailored specifically for molecule generation. Through the reconstruction of over 100 million molecular SELFIES, MolGen internalizes profound structural and grammatical insights. This is further enhanced by domain-agnostic molecular prefix tuning, fostering robust knowledge transfer across diverse domains. Importantly, our self-feedback paradigm steers the model away from ``molecular hallucinations'', ensuring alignment between the model's estimated probabilities and real-world chemical preferences. Extensive experiments on well-known benchmarks underscore MolGen's optimization capabilities in properties such as penalized logP, QED, and molecular docking. Additional analyses affirm its proficiency in accurately capturing molecule distributions, discerning intricate structural patterns, and efficiently exploring the chemical space. Code is available at https://github.com/zjunlp/MolGen.

Steady-state visual evoked potentials (SSVEPs) based brain-computer interface (BCI) has received considerable attention due to its high information transfer rate (ITR) and available quantity of targets. However, the performance of frequency identification methods heavily hinges on the amount of user calibration data and data length, which hinders the deployment in real-world applications. Recently, generative adversarial networks (GANs)-based data generation methods have been widely adopted to create synthetic electroencephalography (EEG) data, holds promise to address these issues. In this paper, we proposed a GAN-based end-to-end signal transformation network for Time-window length Extension, termed as TEGAN. TEGAN transforms short-length SSVEP signals into long-length artificial SSVEP signals. By incorporating a novel U-Net generator architecture and an auxiliary classifier into the network architecture, the TEGAN could produce conditioned features in the synthetic data. Additionally, we introduced a two-stage training strategy and the LeCam-divergence regularization term to regularize the training process of GAN during the network implementation. The proposed TEGAN was evaluated on two public SSVEP datasets (a 4-class dataset and a 12-class dataset). With the assistance of TEGAN, the performance of traditional frequency recognition methods and deep learning-based methods have been significantly improved under limited calibration data. And the classification performance gap of various frequency recognition methods has been narrowed. This study substantiates the feasibility of the proposed method to extend the data length for short-time SSVEP signals for developing a high-performance BCI system. The proposed GAN-based methods have the great potential of shortening the calibration time and cutting down the budget for various real-world BCI-based applications.

The availability of reliable, high-resolution climate and weather data is important to inform long-term decisions on climate adaptation and mitigation and to guide rapid responses to extreme events. Forecasting models are limited by computational costs and, therefore, often generate coarse-resolution predictions. Statistical downscaling, including super-resolution methods from deep learning, can provide an efficient method of upsampling low-resolution data. However, despite achieving visually compelling results in some cases, such models frequently violate conservation laws when predicting physical variables. In order to conserve physical quantities, here we introduce methods that guarantee statistical constraints are satisfied by a deep learning downscaling model while also improving their performance according to traditional metrics. We compare different constraining approaches and demonstrate their applicability across different neural architectures as well as a variety of climate and weather datasets. Besides enabling faster and more accurate climate predictions through downscaling, we also show that our novel methodologies can improve super-resolution for satellite data and standard datasets.

The vast majority of techniques to train fair models require access to the protected attribute (e.g., race, gender), either at train time or in production. However, in many important applications this protected attribute is largely unavailable. In this paper, we develop methods for measuring and reducing fairness violations in a setting with limited access to protected attribute labels. Specifically, we assume access to protected attribute labels on a small subset of the dataset of interest, but only probabilistic estimates of protected attribute labels (e.g., via Bayesian Improved Surname Geocoding) for the rest of the dataset. With this setting in mind, we propose a method to estimate bounds on common fairness metrics for an existing model, as well as a method for training a model to limit fairness violations by solving a constrained non-convex optimization problem. Unlike similar existing approaches, our methods take advantage of contextual information -- specifically, the relationships between a model's predictions and the probabilistic prediction of protected attributes, given the true protected attribute, and vice versa -- to provide tighter bounds on the true disparity. We provide an empirical illustration of our methods using voting data. First, we show our measurement method can bound the true disparity up to 5.5x tighter than previous methods in these applications. Then, we demonstrate that our training technique effectively reduces disparity while incurring lesser fairness-accuracy trade-offs than other fair optimization methods with limited access to protected attributes.

Modern successes of diffusion models in learning complex, high-dimensional data distributions are attributed, in part, to their capability to construct diffusion processes with analytic transition kernels and score functions. The tractability results in a simulation-free framework with stable regression losses, from which reversed, generative processes can be learned at scale. However, when data is confined to a constrained set as opposed to a standard Euclidean space, these desirable characteristics appear to be lost based on prior attempts. In this work, we propose Mirror Diffusion Models (MDM), a new class of diffusion models that generate data on convex constrained sets without losing any tractability. This is achieved by learning diffusion processes in a dual space constructed from a mirror map, which, crucially, is a standard Euclidean space. We derive efficient computation of mirror maps for popular constrained sets, such as simplices and $\ell_2$-balls, showing significantly improved performance of MDM over existing methods. For safety and privacy purposes, we also explore constrained sets as a new mechanism to embed invisible but quantitative information (i.e., watermarks) in generated data, for which MDM serves as a compelling approach. Our work brings new algorithmic opportunities for learning tractable diffusion on complex domains.

Statistical divergences quantify the difference between probability distributions, thereby allowing for multiple uses in machine-learning. However, a fundamental challenge of these quantities is their estimation from empirical samples since the underlying distributions of the data are usually unknown. In this work, we propose a divergence inspired by the Jensen-Shannon divergence which avoids the estimation of the probability density functions. Our approach embeds the data in an reproducing kernel Hilbert space (RKHS) where we associate data distributions with uncentered covariance operators in this representation space. Therefore, we name this measure the representation Jensen-Shannon divergence (RJSD). We provide an estimator from empirical covariance matrices by explicitly mapping the data to an RKHS using Fourier features. This estimator is flexible, scalable, differentiable, and suitable for minibatch-based optimization problems. Additionally, we provide an estimator based on kernel matrices without an explicit mapping to the RKHS. We provide consistency convergence results for the proposed estimator. Moreover, we demonstrate that this quantity is a lower bound on the Jensen-Shannon divergence, leading to a variational approach to estimate it with theoretical guarantees. We leverage the proposed divergence to train generative networks, where our method mitigates mode collapse and encourages samples diversity. Additionally, RJSD surpasses other state-of-the-art techniques in multiple two-sample testing problems, demonstrating superior performance and reliability in discriminating between distributions.

To this end, vast amounts of data generated by various sources, including sensors and other monitoring systems, need to be effectively structured and represented in a way that can be easily understood and processed by both Artificial Intelligence (AI) systems and humans. The digitalisation of the wind energy sector is one of the key drivers for reducing costs and risks over the whole wind energy project life cycle [2]. The digitalisation process encompasses solutions such as digital twins, decision support systems and AI systems, some of which need to still be developed, in order to contribute to reducing operation and maintenance costs, for increasing the amount of energy delivered, as well as for maximising the efficiency of wind energy systems. In this context, the term Knowledge-Based Systems (KBS) refers to AI systems that formalize knowledge as rules, logical expressions, and conceptualisations [3, 4]. Such systems can be realised as AI-enabled digital twins or decision support systems that rely on databases of knowledge (also referred to as knowledge bases or knowledge graphs), which contain machine-readable facts, rules, and logics about a domain of interest, to assist with problem-solving and decision-making [5].

Recent advancements in artificial intelligence (AI) are fundamentally reshaping computing, with large language models (LLMs) now effectively being able to generate and interpret source code and natural language instructions. These emergent capabilities have sparked urgent questions in the computing education community around how educators should adapt their pedagogy to address the challenges and to leverage the opportunities presented by this new technology. In this working group report, we undertake a comprehensive exploration of LLMs in the context of computing education and make five significant contributions. First, we provide a detailed review of the literature on LLMs in computing education and synthesise findings from 71 primary articles. Second, we report the findings of a survey of computing students and instructors from across 20 countries, capturing prevailing attitudes towards LLMs and their use in computing education contexts. Third, to understand how pedagogy is already changing, we offer insights collected from in-depth interviews with 22 computing educators from five continents who have already adapted their curricula and assessments. Fourth, we use the ACM Code of Ethics to frame a discussion of ethical issues raised by the use of large language models in computing education, and we provide concrete advice for policy makers, educators, and students. Finally, we benchmark the performance of LLMs on various computing education datasets, and highlight the extent to which the capabilities of current models are rapidly improving. Our aim is that this report will serve as a focal point for both researchers and practitioners who are exploring, adapting, using, and evaluating LLMs and LLM-based tools in computing classrooms.

How humans and machines make sense of current inputs for relation reasoning and question-answering while putting the perceived information into context of our past memories, has been a challenging conundrum in cognitive science and artificial intelligence. Inspired by human brain's memory system and cognitive architectures, we propose a PMI framework that consists of perception, memory and inference components. Notably, the memory module comprises working and long-term memory, with the latter endowed with a higher-order structure to retain more accumulated knowledge and experiences. Through a differentiable competitive write access, current perceptions update working memory, which is later merged with long-term memory via outer product associations, averting memory overflow and minimizing information conflicts. In the inference module, relevant information is retrieved from two separate memory origins and associatively integrated to attain a more comprehensive and precise interpretation of current perceptions. We exploratively apply our PMI to improve prevailing Transformers and CNN models on question-answering tasks like bAbI-20k and Sort-of-CLEVR datasets, as well as relation calculation and image classification tasks, and in each case, our PMI enhancements consistently outshine their original counterparts significantly. Visualization analyses reveal that memory consolidation, along with the interaction and integration of information from diverse memory sources, substantially contributes to the model effectiveness on inference tasks.