AtomSD Framework for Procedural Atom-Guided Self-Distillation in LLM Agents

AtomSD decomposes procedural knowledge into Intent, When, Do, and Avoid atoms to improve agent reasoning. The self-distillation framework extracts these fine-grained units from successful trajectories, outperforming state-of-the-art baselines by up to 9.31 points.