Virtual Reality (VR) games require players to translate high-level semantic actions into precise device manipulations using controllers and head-mounted displays (HMDs). While humans intuitively perform this translation based on common sense and embodied understanding, whether Large Language Models (LLMs) can effectively replicate this ability remains underexplored. This paper introduces a benchmark, ComboBench, evaluating LLMs' capability to translate semantic actions into VR device manipulation sequences across 262 scenarios from four popular VR games: Half-Life: Alyx, Into the Radius, Moss: Book II, and Vivecraft. We evaluate seven LLMs, including GPT-3.5, GPT-4, GPT-4o, Gemini-1.5-Pro, LLaMA-3-8B, Mixtral-8x7B, and GLM-4-Flash, compared against annotated ground truth and human performance. Our results reveal that while top-performing models like Gemini-1.5-Pro demonstrate strong task decomposition capabilities, they still struggle with procedural reasoning and spatial understanding compared to humans. Performance varies significantly across games, suggesting sensitivity to interaction complexity. Few-shot examples substantially improve performance, indicating potential for targeted enhancement of LLMs' VR manipulation capabilities. We release all materials at https://sites.google.com/view/combobench.

Designing sequences that satisfy multiple, often conflicting, objectives is a central challenge in therapeutic and biomolecular engineering. Existing generative frameworks largely operate in continuous spaces with single-objective guidance, while discrete approaches lack guarantees for multi-objective Pareto optimality. We introduce AReUReDi (Annealed Rectified Updates for Refining Discrete Flows), a discrete optimization algorithm with theoretical guarantees of convergence to the Pareto front. Building on Rectified Discrete Flows (ReDi), AReUReDi combines Tchebycheff scalarization, locally balanced proposals, and annealed Metropolis-Hastings updates to bias sampling toward Pareto-optimal states while preserving distributional invariance. Applied to peptide and SMILES sequence design, AReUReDi simultaneously optimizes up to five therapeutic properties (including affinity, solubility, hemolysis, half-life, and non-fouling) and outperforms both evolutionary and diffusion-based baselines. These results establish AReUReDi as a powerful, sequence-based framework for multi-property biomolecule generation.

This paper examines the generalization capacity of two state-of-the-art classification and similarity learning models in reliably identifying users based on their motions in various Extended Reality (XR) applications. We developed a novel dataset containing a wide range of motion data from 49 users in five different XR applications: four XR games with distinct tasks and action patterns, and an additional social XR application with no predefined task sets. The dataset is used to evaluate the performance and, in particular, the generalization capacity of the two models across applications. Our results indicate that while the models can accurately identify individuals within the same application, their ability to identify users across different XR applications remains limited. Overall, our results provide insight into current models generalization capabilities and suitability as biometric methods for user verification and identification. The results also serve as a much-needed risk assessment of hazardous and unwanted user identification in XR and Metaverse applications. Our cross-application XR motion dataset and code are made available to the public to encourage similar research on the generalization of motion-based user identification in typical Metaverse application use cases.

Prior literature has explored the use of fundamental information as a proxy for firm linkages. If investors have limited attention, then news impacting the price of a firm may only slowly be incorporated into prices of related firms, leading to return predictability across firms. Indeed, for many such firm linkages it has been shown that lagged returns of a firm are predictive of future returns for firms which are more similar to it. This effect is sometimes referred to as a momentum spillover effect, or a lead-lag strategy. Momentum spillover has been documented for similarities formed from a variety of fundamental information including industry [24], supply chain [12], analyst coverage [1], and geography [32], among others. Unrelated literature explores the application of machine learning techniques to the learning of similarity relations between securities, often with the goal of clustering securities for risk management, signal generation, or portfolio construction. See e.g. the literature review in [37] for examples of classification and clustering techniques, [44] for a demonstration of how embeddings from Large Language Models can be used to extract company similarity relations, or [6] for a more general review of machine learning applications in finance. More recent work has begun to explore the use of supervised learning techniques to extract similarity relationships.

Outdated facts in temporal knowledge graphs (TKGs) result from exceeding the expiration date of facts, which negatively impact reasoning performance on TKGs. However, existing reasoning methods primarily focus on positive importance of historical facts, neglecting adverse effects of outdated facts. Besides, training on these outdated facts yields extra computational cost. To address these challenges, we propose an outdated fact filtering framework named HALO, which quantifies the temporal validity of historical facts by exploring the half-life theory to filter outdated facts in TKGs. HALO consists of three modules: the temporal fact attention module, the dynamic relation-aware encoder module, and the outdated fact filtering module. Firstly, the temporal fact attention module captures the evolution of historical facts over time to identify relevant facts. Secondly, the dynamic relation-aware encoder module is designed for efficiently predicting the half life of each fact. Finally, we construct a time decay function based on the half-life theory to quantify the temporal validity of facts and filter outdated facts. Experimental results show that HALO outperforms the state-of-the-art TKG reasoning methods on three public datasets, demonstrating its effectiveness in detecting and filtering outdated facts (Codes are available at https://github.com/yushuowiki/K-Half/tree/main ).

Global climate change has reduced crop resilience and pesticide efficacy, making reliance on synthetic pesticides inevitable, even though their widespread use poses significant health and environmental risks. While these pesticides remain a key tool in pest management, previous machine-learning applications in pesticide and agriculture have focused on classification or regression, leaving the fundamental challenge of generating new molecular structures or designing novel candidates unaddressed. In this paper, we propose Pesti-Gen, a novel generative model based on variational auto-encoders, designed to create pesticide candidates with optimized properties for the first time. Specifically, Pesti-Gen leverages a two-stage learning process: an initial pre-training phase that captures a generalized chemical structure representation, followed by a fine-tuning stage that incorporates toxicity-specific information. The model simultaneously optimizes over multiple toxicity metrics, such as (1) livestock toxicity and (2) aqua toxicity to generate environmentally friendly pesticide candidates. Notably, Pesti-Gen achieves approximately 68\% structural validity in generating new molecular structures, demonstrating the model's effectiveness in producing optimized and feasible pesticide candidates, thereby providing a new way for safer and more sustainable pest management solutions.

Monitoring the behavior of automated real-time stream processing systems has become one of the most relevant problems in real world applications. Such systems have grown in complexity relying heavily on high dimensional input data, and data hungry Machine Learning (ML) algorithms. We propose a flexible system, Feature Monitoring (FM), that detects data drifts in such data sets, with a small and constant memory footprint and a small computational cost in streaming applications. The method is based on a multi-variate statistical test and is data driven by design (full reference distributions are estimated from the data). It monitors all features that are used by the system, while providing an interpretable features ranking whenever an alarm occurs (to aid in root cause analysis). The computational and memory lightness of the system results from the use of Exponential Moving Histograms. In our experimental study, we analyze the system's behavior with its parameters and, more importantly, show examples where it detects problems that are not directly related to a single feature. This illustrates how FM eliminates the need to add custom signals to detect specific types of problems and that monitoring the available space of features is often enough.

Gaming never went out of style, but in 2020, it evolved from a fun hobby into an essential lifeline. Staying sane isn't easy when you're stuck in isolation for months on end. You can only watch so much Netflix before your brain starts dripping out of your ears. Games provide more active experiences that can help you forget that you've been staring at the same walls for weeks, letting you explore far-away virtual worlds or hang out with friends in multiplayer lobbies. In 2020, gaming became vital.

The ongoing covid-19 pandemic placed a brighter-than-usual spotlight on gaming in 2020, with an isolated population looking for entertainment they could enjoy from the safety of home. How fortunate then that alongside the year's many maladies, 2020 also delivered some of the most memorable games in recent years. From laid-back life simulators to an anticipated sequel that scrutinized cyclical violence, the gaming world was replete with options for anyone who wanted to get their minds off the consistently grim reality around them. The reintroduction and reimagination of the classic "Final Fantasy VII" highlighted the early spring, while the November debut of the PlayStation 5 ushered in a next-generation hero the gaming world both needed and deserved. Even with multiple delays pushing the much-anticipated "Cyberpunk 2077" beyond our Dec. 1 cutoff for Game of The Year consideration, there was no shortage of worthy contenders for that title.

Summer and fall are often the worst time to be a video game fan. Publishers often hold their best stuff til the end of the year, and it's worse this year, because Microsoft and Sony are hanging on to their biggest games until the Xbox Series X and PlayStation 5 are out this holiday season. Still, 2020 has already offered an embarrassment of riches for gamers. Here are the best video games of 2020 so far, to tide you over til year's end: Doom Eternal ripped and tore its way into our hearts at the beginning of the year and hasn't been topped since. What makes Doom Eternal so remarkable is that it managed to improve on its predecessor, and in so doing proved the almost 30-year-old franchise is still as vibrant and vital as it was in 1993.