MEMIG: Self-evolving Memory Construction for Antibody Optimization

Memory has become central to self-evolving LLM agents, where they store and retrieve experience from past rollouts to guide future decisions. Existing systems represent this experience as free-form natural language and maintain it by rewriting, which fits tasks whose experience is narratable. However, in scientific optimization, experience is a quantity measured by an oracle under a specific context. Rewriting it as text replaces a measurement with a description, leaves no index for where a record should be, and cannot express how interventions interact. To address this, we argue that agent memory for scientific optimization should instead be indexable, quantitative, and structured, and we introduce MEMIG, a memory construction framework that realizes these three properties. In detail, MEMIG addresses each insight by a domain-grounded identifier and accumulates the quantitative effect measured by a property oracle, so that updates remain local and retrieval becomes an exact lookup. Next, MEMIG connects insights whose interventions are applied together and measures their epistasis by decomposing each evaluated set. Our results show that MEMIG improves agentic performance in majority of settings across in silico antibody optimization tasks.