Recent advancements in large language models (LLMs) underscore the need for more comprehensive evaluation methods to accurately assess their reasoning capabilities. Existing benchmarks are often domain-specific and thus cannot fully capture an LLM’s general reasoning potential. To address this limitation, we introduce the **Knowledge Orthogonal Reasoning Gymnasium (KORGym)**, a dynamic evaluation platform inspired by KOR-Bench and Gymnasium. KORGym offers over fifty games in either textual or visual formats and supports interactive, multi-turn assessments with reinforcement learning scenarios. Using KORGym, we conduct extensive experiments on 19 LLMs and 8 VLMs, revealing consistent reasoning patterns within model families and demonstrating the superior performance of closed-source models. Further analysis examines the effects of modality, reasoning strategies, reinforcement learning techniques, and response length on model performance. We expect KORGym to become a valuable resource for advancing LLM reasoning research and developing evaluation methodologies suited to complex, interactive environments.

Abstract In this article, we propose a novel hybrid framework by combining smoothed particle hydrodynamics and adaptive narrow band fluid implicit particle method (NB‐FLIP) to faithfully model the multiphysical processes involving heat transfer and phase transition, and to precisely simulate the dynamics of condensed droplets moving along intricate objects. We first formulate a governing physical model built upon an improved phase transition model and an augmented on‐surface drop analysis method to achieve realistic condensation effects over intricate hydrophilic/hydrophobic interface. To achieve both high‐fidelity interactions and high‐resolution visual effects, we further develop an adaptive NB‐FLIP solver with octree‐dictated background grid in order to further enhance the performance of our framework. Experimental results have shown that our approach can be used to efficiently and realistically simulate the small‐scale interaction details between condensed drops and complex objects with arbitrary geometry.

Mobile edge computing (MEC), which enables smart terminals to actively offload computation workloads to computational servers deployed at the edge of networks, has provided an efficient approach to address the intensive computation requirement in mobile Internet applications. In this paper, we investigate the delay-aware computation offloading via MEC for Internet of Things (IoT) with secrecy provisioning. Specifically, we consider a scenario where a malicious eavesdropper intentionally overhears the IoT devices’ offloaded computational data. Taking into account the secrecy outage due to the eavesdropper’s overhearing, we formulate a joint optimization of the secrecy-provisioning, computation offloading, and radio resource allocation (including time and power allocations), with the objective of minimizing the overall delay in finishing the computation requirement of the IoT device. Despite the nonconvexity of the joint optimization problem, we propose an efficient algorithm to compute the optimal computation offloading solution. By exploiting the optimal offloading decision of each IoT device, we further consider the scenario of a group of IoT devices offloading computation workloads to the edge server, and investigate how the edge server optimally selects the devices for providing the computation offloading service while subject to the limited energy budget and the time-slot budget of the edge server. We propose an efficient algorithm to find the optimal selection of the devices. We present extensive numerical results to validate the effectiveness of our proposed algorithms and show the impact of the secrecy requirement.