As an emerging non-traditional security threat, cognitive warfare undermines social systems by manipulating information dissemination, distorting cognitive consensus, and eroding social trust. The reliability of social systems against cognitive warfare reflects their ability to maintain core functions, resist cognitive infiltration, and restore social order. To quantitatively evaluate this reliability, this paper proposes a hybrid approach integrating the Extended Fault Tree (EFT) and Multi-Valued Decision Diagram (MDD). First, key failure paths and interdependencies are identified, including cascading failures among four core subsystems: information dissemination, cognitive consensus, socia... More >

The ultra-large scale and prolonged runtime of Large Language Model (LLM) training—often involving thousands of GPUs and spanning weeks—render reliability a pivotal bottleneck. Hardware failures, stragglers, and runtime issues can waste over 30% of GPU resources, delaying model rollout and driving up costs. This survey focuses on reliability optimization for LLM training systems. The discussion centers on three pillars of reliability: fault detection, fault recovery, and straggler mitigation. For each pillar, we dissect innovative mechanisms, which range from communication-aware fault detection to adaptive load balancing, and we assess their impact on critical reliability metrics such as... More >