Search for a command to run...
Multi-agent systems increasingly require transparency alongside performance optimization. This chapter presents ARES-SC, integrating blockchain with multi-agent reinforcement learning to address trust-performance trade-offs. The system combines (1) adaptive resolution for dynamic resource allocation, (2) semantic communication for interpretable coordination, and (3) blockchain for immutable decision recording. Experimental validation across three scenarios reveals important trade-offs. ARES-CORE achieved highest performance (normalized scores 2900-3500) but with elevated overhead (6657-6875%). ARES-BC provided competitive performance (2500-3000) with complete cryptographic verification at measurable costs (625-737 blockchain units). All methods exhibited significant computational overhead (6200-6800%). This work provides honest assessment of when transparency features justify computational costs, offering practical guidance for trustworthy multi-agent systems.