ARM: Role-Conditioned Neuron Transplantation for Training-Free Generalist LLM Agent Merging
This addresses the need for generalist LLM agents that can robustly adapt across interactive environments, though it is incremental as it builds on existing model merging techniques.
The paper tackles the problem of specialized large language model agents failing to adapt to other environments by proposing ARM, a training-free model merging method that improves cross-benchmark generalization and outperforms prior merging methods and domain-specific experts across diverse domains.
Interactive large language model agents have advanced rapidly, but most remain specialized to a single environment and fail to adapt robustly to other environments. Model merging offers a training-free alternative by integrating multiple experts into a single model. In this paper, we propose Agent-Role Merging (ARM), an activation-guided, role-conditioned neuron transplantation method for model merging in LLM agents. ARM improves existing merging methods from static natural language tasks to multi-turn agent scenarios, and over the generalization ability across various interactive environments. This is achieved with a well designed 3-step framework: 1) constructing merged backbones, 2) selection based on its role-conditioned activation analysis, and 3) neuron transplantation for fine-grained refinements. Without gradient-based optimization, ARM improves cross-benchmark generalization while enjoying efficiency. Across diverse domains, the model obtained via ARM merging outperforms prior model merging methods and domain-specific expert models, while demonstrating strong out-of-domain generalization.