Compositional Safety Failures in Harness Evolution: Identification and Runtime Monitoring

arXiv:2609.33123 · cs.AI, cs.SE · Submitted 2026-09-27 · Read on arXiv

cs.AI, cs.SE

Submitted: 2026-09-27

Updated: 2026-09-27

Terminology

• Hypergraph — 41× in this paper · explained in 2 episodes
• Benchmark — 22× in this paper · explained in 2 episodes
• Verification — 17× in this paper · explained in 2 episodes
• Compositional Failures — 14× in this paper
• Provenance — 8× in this paper · explained in 4 episodes
• Large Language Model — 5× in this paper
• Self-Evolving Agents — 5× in this paper · explained in 2 episodes
• Agent Harness — 3× in this paper · explained in 2 episodes
• Abstraction — 2× in this paper · explained in 1 episode
• Adaptation — 2× in this paper · explained in 1 episode
• Compositional Failure Rate — 2× in this paper
• Generative AI — 2× in this paper · explained in 2 episodes
• Reinforcement Learning — 2× in this paper · explained in 4 episodes
• Trace — 2× in this paper · explained in 2 episodes
• AI Agents — 1× in this paper · explained in 2 episodes
• Agentic Reasoning — 1× in this paper · explained in 1 episode
• Alignment — 1× in this paper · explained in 1 episode
• Benchmarking — 1× in this paper · explained in 3 episodes
• Chain-of-Thought Prompting — 1× in this paper · explained in 2 episodes
• Foundation Model — 1× in this paper · explained in 3 episodes
• Foundation Models — 1× in this paper · explained in 7 episodes
• Grounding — 1× in this paper · explained in 3 episodes
• Large Language Model (LLM) — 1× in this paper · explained in 3 episodes
• Prompt Engineering — 1× in this paper · explained in 4 episodes

Sources

• A Survey of Self-Evolving Agents: What, When, How, and Where to Evolve on the Path to Artificial Super Intelligence
• Meta-Harness: End-to-End Optimization of Model Harnesses
• Position: Agentic Evolution is the Path to Evolving LLMs
• Safety in Self-Evolving LLM Agent Systems: Threats, Amplification, and Case Studies
• SafeEvolve: Harness-Policy Co-Evolution from Agent Experience for Safety Alignment
• TTHE: Test-Time Harness Evolution
• Natural-Language Agent Harnesses
• Alita: Generalist Agent Enabling Scalable Agentic Reasoning with Minimal Predefinition and Maximal Self-Evolution
• SHE: Trajectory-driven Safety Harness Evolution for LLM Agents
• Toolformer: Language Models Can Teach Themselves to Use Tools
• The Prompt Report: A Systematic Survey of Prompt Engineering Techniques
• When Self-Evolution Backfires: Pre-Commit Gating against Skill Contamination in LLM Agents
• Reflexion: Language Agents with Verbal Reinforcement Learning
• WikiSkill: Compiling Agent Experience into Persistent Knowledge for Skill Evolution
• Voyager: An Open-Ended Embodied Agent with Large Language Models
• OEP: Poisoning Self-Evolving LLM Agents via Locally Correct but Non-Transferable Experiences
• Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
• Verify Smarter, Evolve Further: Efficient Harness Evolution through Behavior-Aware Verification
• Benign Alone, Harmful Together: Exploiting Experience Composition in Self-Evolving LLM Agents

Related papers

• MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
• Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
• The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
• MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
• Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems
• VSAL: A Vision Solver with Adaptive Layouts for Graph Property Detection