AutoSpec: Safety Rule Evolution for LLM Agents via Inductive Logic Programming
Pingchuan Ma, Zhaoyu Wang, Zimo Ji, Yuguang Zhou, Zhantong Xue, Zongjie Li, Shuai Wang, Xiaoqin Zhang
Why It Matters
What makes this one worth your time
As LLM agents become more autonomous, ensuring their safety through interpretable and effective rules is crucial for preventing harmful actions and data leaks.
AutoSpec refines safety rules for LLM agents using ILP to balance precision and recall.
Summary
The paper presents AutoSpec, a framework that evolves safety rules for large language model agents using inductive logic programming and counterexample-guided inductive synthesis. It aims to improve the balance between precision and recall of safety rules by iteratively refining them based on user annotations of safe and unsafe behaviors.
Key contributions
- Development of AutoSpec, a framework for evolving safety rules using ILP and CEGIS.
- Demonstration of significant improvements in rule precision and recall across multiple domains.
- Provision of human-readable and auditable safety rules that generalize to unseen scenarios.
Notable insights
- Using ILP to identify predicates that frequently appear in false negatives but rarely in false positives helps prune the search space for rule edits.
- The iterative approach of AutoSpec allows for continuous improvement of safety rules based on real-world annotations.
Possible limitations
- Not stated in the abstract
Abstract
arXiv:2606.24245v2 Announce Type: cross Abstract: Large language model (LLM) agents increasingly automate complex tasks by integrating language models with external tools and environments. However, their autonomy poses significant safety risks: agents may execute destructive commands, leak sensitive data, or violate domain constraints. Existing safety approaches face a fundamental tradeoff: hand-crafted rules are interpretable but brittle, with overly conservative rules blocking safe operations (high false positives) while permissive rules miss unsafe behaviors (high false negatives). Neural classifiers lack the interpretability required for safety-critical deployments. We present AutoSpec, a framework that automatically evolves deployed expert-designed safety rules from user safe/unsafe annotations through counterexample-guided inductive synthesis (CEGIS) guided by inductive logic programming (ILP). Starting from the expert rules and a stream of annotated traces, AutoSpec iteratively evaluates rules, mines false-positive and false-negative counterexamples, uses ILP to learn which predicates discriminate them, generates candidate rule edits, and verifies candidates to select the best revision. The key insight is that ILP efficiently identifies predicates that appear frequently in false negatives but rarely in false positives (or vice versa), dramatically pruning the exponential search space of rule edits. This continues until convergence, producing interpretable rules that balance precision and recall. We evaluate AutoSpec on 291 execution traces spanning code execution and embodied agent domains. AutoSpec raises rule F1 to 0.98 and 0.93 across the two domains, achieving up to 94% false positive reduction while maintaining high recall, and converges within 4-5 iterations. The ILP-guided approach achieves up to 4.8x higher F1 than heuristic CEGIS. The learned rules are human-readable, auditable, and generalize to unseen scenarios.