Phantom Guardrails: When Self-Improving Agent Harnesses Fix Failures That Never Happened

arXiv:2607.13083 · cs.CR, cs.SE · Submitted 2026-07-13 · Read on arXiv

Su Wang, Pin Qian, Yifan Lin, Jingzhou Xu, Yihang Chen, Xiaochong Jiang, Lifei Liu, Haoran Yu

cs.CR, cs.SE

Submitted: 2026-07-13

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Reward Hacking — 7× in this paper · explained in 8 episodes
• Abstention — 6× in this paper · explained in 1 episode
• Steering — 6× in this paper · explained in 2 episodes
• Trace — 5× in this paper · explained in 2 episodes
• Benchmark — 4× in this paper · explained in 2 episodes
• Pruning — 4× in this paper · explained in 4 episodes
• Provenance — 3× in this paper · explained in 4 episodes
• AI Agents — 2× in this paper · explained in 2 episodes
• Accuracy — 2× in this paper · explained in 1 episode
• Alignment — 2× in this paper · explained in 1 episode
• LLM-as-a-Judge — 2× in this paper · explained in 3 episodes
• Parsimony — 2× in this paper · explained in 2 episodes
• Scaffolding — 2× in this paper · explained in 3 episodes
• Adam — 1× in this paper · explained in 2 episodes
• Adaptation — 1× in this paper · explained in 1 episode
• Agent Harness — 1× in this paper · explained in 2 episodes
• Agentic AI — 1× in this paper · explained in 1 episode
• Benchmarking — 1× in this paper · explained in 3 episodes
• Failure Taxonomy — 1× in this paper · explained in 2 episodes
• Fine-Tuning — 1× in this paper · explained in 5 episodes
• LLM judge — 1× in this paper · explained in 2 episodes
• Reinforcement Learning — 1× in this paper · explained in 4 episodes

Sources

• QiMeng-PRepair: Precise Code Repair via Edit-Aware Reward Optimization
• Meta-Harness: End-to-End Optimization of Model Harnesses
• Analysing Chain of Thought Dynamics: Active Guidance or Unfaithful Post-hoc Rationalisation?
• SafetyRepro: Configuration-Conditional Rank Instability on Alignment Benchmarks
• Agentic Harness Engineering: Observability-Driven Automatic Evolution of Coding-Agent Harnesses
• Reflect-Guard: Enhancing LLM Safeguards against Adversarial Prompts via Logical Self-Reflection
• You Cannot Fix What You Cannot Find! An Investigation of Fault Localization Bias in Benchmarking Automated Program Repair Systems
• AutoHarness: improving LLM agents by automatically synthesizing a code harness
• Categorizing Variants of Goodhart's Law
• Code as Agent Harness
• Evolving Agents in the Dark: Retrospective Harness Optimization via Self-Preference
• Defining and Characterizing Reward Hacking
• LLM Agents Already Know When to Call Tools -- Even Without Reasoning
• Beyond Accuracy: Measuring Bias Acknowledgment in Chain-of-Thought Reasoning for Responsible AI Evaluation
• VeRO: A Harness for Agents to Optimize Agents
• The Observability Gap: Why Output-Level Human Feedback Fails for LLM Coding Agents
• Adapting the Interface, Not the Model: Runtime Harness Adaptation for Deterministic LLM Agents
• PAFT: Preservation Aware Fine-Tuning for Minimal-Edit Program Repair
• Reward Auditor: Inference on Reward Modeling Suitability in Real-World Perturbed Scenarios
• OR-Bench: An Over-Refusal Benchmark for Large Language Models

Related papers

• SoK: AI-Augmented Binary Reversing
• Relaxed Sender Anonymity for CBDC Interbank Settlement: A Zero-Knowledge Approach on Permissioned EVM
• Calibration-Family Overfit: Why Trusted Sabotage Monitors Don't Transfer Across Lineages
• Efficient Fuzzy PSI under One-Sided Assumptions
• Sealing the Audit-Runtime Gap for LLM Skills
• Token Composition: A Graph Based on EVM Logs