Procedural Graphs: Self-Evolving Execution Structures for LLM Agents
Summary
Procedural Graphs keep what-to-do knowledge outside model weights as an editable directed graph of procedure nodes and attributed transitions.
Mirror knowledge graphs for procedures.
Store (procedure, relation, procedure) triplets so agents answer what-to-do with an inspectable graph instead of reconstructing order from a flat history.
a Procedural Graph organizes procedural knowledge into (procedure, relation, procedure) triplets for what-to-do questions.
Guide from the local neighborhood.
Locate the active node, retrieve its k-hop transitions with condition/guidance/pitfalls attributes, and generate step-level situational advice rather than dumping the full graph.
Bias without dictating.
Append generative guidance to the solver prompt so the ReAct loop stays free to reason while admissible transitions stay explicit.
biases the solver’s next action without dictating it.
Evolve under a validation gate.
Contrast failed versus successful trajectories, edit topology and attributes, keep only candidates that do not hurt held-out score, and retain rejected edits as negative constraints.
The graph is self-evolving: an LLM refiner contrasts failed trajectories with successful ones
Construction modes matter.
Evolution from scratch can match or surpass expert graphs; iterative repair recovers when a hand-crafted prior initially hurts performance.
Starting from a minimal skeleton, the loop builds graphs that match or surpass hand-designed ones.
Read the original ↗Save to your Sponge
Summarized with Sponge. Quotes are the author’s own words.