Weili Wang

Ju Yang, Weili Wang, Jianyu Niu et al.

Confidential blockchains leveraging Trusted Execution Environments (TEEs) have garnered extensive attention for transaction confidentiality. In this paper, we first taxonomize two classes of attacks against confidential blockchains, i.e., execution-inference and execution-replay attacks, which exploit TEEs' long-lasting side-channel and state-continuity issues to compromise the confidentiality of existing consortium blockchains. Then, we present ODYSSEY, a confidential blockchain that efficiently mitigates these attacks. The core innovations of ODYSSEY are the following: (1) Its delegation model: clients delegate transaction execution to their designated trustees, while other participants synchronize only the execution results, which significantly reduces the attack surface while preserving confidentiality and system performance. (2) Two novel techniques to improve ODYSSEY's efficiency and security: location-aware concurrent execution and delegation failure handler. Finally, we develop a prototype of ODYSSEY on FISCO BCOS, an enterprise-grade consortium blockchain platform. We have conducted various experiments, and our evaluation results show that in a WAN environment with 3 nodes, ODYSSEY can achieve about 4k throughput while keeping latency as low as 0.4-0.5s.

Weili Wang, Omid Abbasi, Halim Yanikomeroglu et al.

Vertical heterogenous networks (VHetNets) and artificial intelligence (AI) play critical roles in 6G and beyond networks. This article presents an AI-native VHetNets architecture to enable the synergy of VHetNets and AI, thereby supporting varieties of AI services while facilitating automatic and intelligent network management. Anomaly detection in Internet of Things (IoT) is a major AI service required by many fields, including intrusion detection, state monitoring, device-activity analysis, security supervision and so on. Conventional anomaly detection technologies mainly consider the anomaly detection as a standalone service that is independent of any other network management functionalities, which cannot be used directly in ubiquitous IoT due to the resource constrained end nodes and decentralized data distribution. In this article, we develop an AI-native VHetNets-enabled framework to provide the anomaly detection service for ubiquitous IoT, whose implementation is assisted by intelligent network management functionalities. We first discuss the possibilities of VHetNets used for distributed AI model training to provide anomaly detection service for ubiquitous IoT, i.e., VHetNets for AI. After that, we study the application of AI approaches in helping provide automatic and intelligent network management functionalities for VHetNets, i.e., AI for VHetNets, whose aim is to facilitate the efficient implementation of anomaly detection service. Finally, a case study is presented to demonstrate the efficiency and effectiveness of the proposed AI-native VHetNets-enabled anomaly detection framework.

Minghao Yan, Bo Peng, Benjamin Coleman et al.

Large Language Models (LLMs) have emerged as powerful operators for evolutionary search, yet the design of efficient search scaffolds remains ad hoc. While promising, current LLM-in-the-loop systems lack a systematic approach to managing the evolutionary process. We identify three distinct failure modes: Context Pollution, where experiment history biases future candidate generation; Mode Collapse, where agents stagnate in local minima due to poor exploration-exploitation balance; and Weak Collaboration, where rigid crossover strategies fail to leverage parallel search trajectories effectively. We introduce Progress-Aware Consistent Evolution (PACEvolve), a framework designed to robustly govern the agent's context and search dynamics, to address these challenges. PACEvolve combines hierarchical context management (HCM) with pruning to address context pollution; momentum-based backtracking (MBB) to escape local minima; and a self-adaptive sampling policy that unifies backtracking and crossover for dynamic search coordination (CE), allowing agents to balance internal refinement with cross-trajectory collaboration. We demonstrate that PACEvolve provides a systematic path to consistent, long-horizon self-improvement, achieving state-of-the-art results on LLM-SR and KernelBench, while discovering solutions surpassing the record on Modded NanoGPT.

Minghao Yan, Bo Peng, Benjamin Coleman et al.

Large language models have become drivers of evolutionary search, but most systems rely on a fixed, prompt-elicited policy to sample next candidates. This limits adaptation in practical engineering and research tasks, where evaluations are expensive, and progress depends on learning task-specific search dynamics. We introduce PACEvolve++, an advisor-model reinforcement learning framework for test-time policy adaptation in evolutionary search agents. PACEvolve++ decouples strategic search decisions from implementation: a trainable advisor generates, assesses, and selects hypotheses, while a stronger frontier model translates selected hypotheses into executable candidates. To train the advisor under non-stationary feedback, we propose a phase-adaptive approach that adapts its optimization strategy to different phases of the evolutionary process. Early in evolution, it uses group-relative feedback to learn broad search preferences; later, as reward gaps compress, it emphasizes best-of-$k$ frontier contribution to support stable refinement. Across expert-parallel load balancing, sequential recommendation, and protein fitness extrapolation, PACEvolve++ outperforms the state-of-the-art evolutionary search framework with frontier models, achieving faster convergence and stabilizing test-time training during evolutionary search.