arXiv Science⌕ Search

arXiv subjects

Jeffrey Chu

Publications and source records attributed to Jeffrey Chu.

2 recordsLinked to original sources

Peer Memory Changes Who LLM Agents Consult Without a Detected Gain in Answer Accuracy

Multi-agent LLM systems often reuse the same agents across tasks, creating an opportunity to draw on past interactions when deciding whom to consult. We study this form of memory with PeerMem, which saves MACE's LinUCB state and reuses it when the same agents return. Using six agents, a fixed set of 80 HotpotQA questions, and 12 paired lineages per model, we compare cold, warm, and shuffled initialization in an answer-exchange protocol with reference-answer feedback. Under stable evidence access, warm memory lowers pooled hindsight regret and increases consultation of other evidence holders on Qwen3-32B and Gemma-4-31B-it, while final token F1 differences remain uncertain. Qwen interaction traces further show that warm agents more often encounter peer answers that score above their own prior answers. We extend the evaluation to Qwen3.5-9B using freshly learned memory to compare warm and cold initialization under reassigned and paired stable roles. The primary changed-minus-stable final-F1 interaction is -1.38 points, with a 95% interval of [-3.50,+0.73]. Together, these experiments trace the effects of peer memory through consultation, answer revision, and final task performance.

cs.MA↗

Enhancing Security in Blockchain Networks: Anomalies, Frauds, and Advanced Detection Techniques

Blockchain technology, a foundational distributed ledger system, enables secure and transparent multi-party transactions. Despite its advantages, blockchain networks are susceptible to anomalies and frauds, posing significant risks to their integrity and security. This paper offers a detailed examination of blockchain's key definitions and properties, alongside a thorough analysis of the various anomalies and frauds that undermine these networks. It describes an array of detection and prevention strategies, encompassing statistical and machine learning methods, game-theoretic solutions, digital forensics, reputation-based systems, and comprehensive risk assessment techniques. Through case studies, we explore practical applications of anomaly and fraud detection in blockchain networks, extracting valuable insights and implications for both current practice and future research. Moreover, we spotlight emerging trends and challenges within the field, proposing directions for future investigation and technological development. Aimed at both practitioners and researchers, this paper seeks to provide a technical, in-depth overview of anomaly and fraud detection within blockchain networks, marking a significant step forward in the search for enhanced network security and reliability.

cs.CR↗