Self-Evolving Agent Harness
cs.AI, cs.CL
Submitted: 2026-10-13
Updated: 2026-10-13
Comments: 18 pages, 3 figures
Code: https://github.com/zgcagi/Harness-RSI
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscation
- MOSS: Self-Evolution through Source-Level Rewriting in Autonomous Agent Systems
- A Theory of Reliable Self-Evolution for Agent Harnesses
- CoTrace: Data Recipes for Training Terminal Agents with Harness-Model Co-Evolution
- StressEval: Failure-Driven Dynamic Benchmarking for Knowledge-Intensive Reasoning in Large Language Models
- Generalization in Adaptive Data Analysis and Holdout Reuse
- Harness Evolution as Learning: Approximation, Generalization, and Optimization Limits of Self-Improving Personal Agents
- SkillMOO: Multi-Objective Optimization of Agent Skills for Software Engineering
- LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code
- MedAgentBench: A Realistic Virtual EHR Environment to Benchmark Medical LLM Agents
- Memory Transfer Planning: LLM-driven Context-Aware Code Adaptation for Robot Manipulation
- EMHO: EMbodied Agent Harness Optimization via Experience Traces
- Meta-Harness: End-to-End Optimization of Model Harnesses
- Argus: A General-Purpose Agentic Reasoning Runtime for Long-Horizon Tasks
- Finding the Right Fit: Model-Harness Interactions across Agent Tasks
- Agentic Harness Engineering: Observability-Driven Automatic Evolution of Coding-Agent Harnesses
- AgentRewardBench: Evaluating Automatic Evaluations of Web Agent Trajectories
- HarnessBank: Semantic Gene-Bank Search with Gated Verification for Agent-Harness Self-Evolution
- SkillClaw: Let Skills Evolve Collectively with Agentic Evolver
- Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills
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
- Agents' Last Exam
- 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