Large language models have made role-playing agents viable as interactive characters, virtual companions, game agents, and social actors. The harder problem, however, is not producing one convincing response but sustaining a coherent role through long, changing, and sometimes adversarial interactions. This survey examines recent work on role consistency: the maintenance of identity, character knowledge, expressive style, memory, values, behavioral logic, and safety boundaries over time. We organize the literature around role specification, role-state maintenance, role-aware reasoning, role-consistent alignment, and evaluation.

Training Large Language Model (LLM) agents via Reinforcement Learning (RL) in federated settings faces fundamental challenges, including model heterogeneity across clients, prohibitive communication costs and severe data heterogeneity. We propose a semantic federated learning framework called SemFed that addresses all challenges by communicating behavioral patterns rather than model parameters. Each client extracts compact semantic policy summaries from local training trajectories, which are aggregated at the server into shared, elite and anti-patterns. These aggregated patterns are injected back into local training through different complementary channels. Since the communicated objects are structured behavioral abstractions, SemFed supports heterogeneous model architectures while reducing per-round communication to kilobytes. Experiments on ALFWorld and WebShop across multiple heterogeneous LLM backbones demonstrate that SemFed consistently surpasses strong federated baselines in task success while reducing communication overhead by orders of magnitude.