Search papers, labs, and topics across Lattice.
The paper introduces SERO, a framework for contract-preserving role evolution in multi-agent LLM systems, addressing the problem of role drift and broken answer contracts when adapting role-based systems. SERO enforces five structural contracts (capability, communication, validation, aggregation, output protocol) during role evolution. Experiments on real-world reasoning benchmarks using multiple LLMs demonstrate that SERO's credit-guided retrieval, communication DAG, and contextual-bandit controller successfully preserve contracts and improve task performance.
Multi-agent LLM systems can adapt to new tasks without sacrificing structural integrity, thanks to a novel framework that guarantees role evolution preserves key operational contracts.
Role-based LLM multi-agent systems need adaptive role pools, yet adapting such systems is not merely a matter of prompt optimization: roles often carry structural obligations, including capability coverage, message compatibility, validation, final-answer aggregation, and parser-compatible output protocols. Existing systems either fix the role inventory and lose adaptivity, or allow unconstrained generation to induce role drift, removing structurally necessary roles and breaking answer contracts. We formulate this as contract-preserving role evolution, requiring every committed edit to preserve five structural contracts (capability, communication, validation, aggregation, output protocol). We instantiate this formulation in SERO, a Self-Evolving Role Orchestration framework that evolves a typed role-card pool through credit-guided retrieval, a credit-ranked communication DAG with a protected terminal aggregator and conditional validator repair, and a contextual-bandit controller whose LLM-proposed edits are committed only when they preserve the contracts and improve task score. Experiments on real-world reasoning benchmarks across three LLM backbones confirm the value of contract-preserving role evolution.
