HarnessMDP: Learning to Control LLM Agent Harnesses with Offline Reinforcement Learning

arXiv:2607.05458 · cs.LG, cs.AI · Submitted 2026-07-05 · Read on arXiv

cs.LG, cs.AI

Submitted: 2026-07-05

Updated: 2026-10-15

Code: https://github.com/Hik289/Agentic-RL-harness

Terminology

• Ablation — 1× in this paper · explained in 1 episode
• Benchmark — 1× in this paper · explained in 2 episodes
• Large Language Model (LLM) — 1× in this paper · explained in 3 episodes
• Reinforcement Learning — 1× in this paper · explained in 4 episodes
• Verification — 1× in this paper · explained in 2 episodes
• Verifier — 1× in this paper · explained in 2 episodes

Sources

• GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning
• DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines
• Process-Level Trajectory Evaluation for Environment Configuration in Software Engineering Agents
• Offline Reinforcement Learning: Tutorial, Review, and Perspectives on Open Problems
• Agentic Harness Engineering: Observability-Driven Automatic Evolution of Coding-Agent Harnesses
• ToolSandbox: A Stateful, Conversational, Interactive Evaluation Benchmark for LLM Tool Use Capabilities
• Evaluating Very Long-Term Conversational Memory of LLM Agents
• GAIA: a benchmark for General AI Assistants
• Evolving Agents in the Dark: Retrospective Harness Optimization via Self-Preference
• Strained Coherence: A Pre-Failure Signal in Coding Agent Execution Trajectories
• Advantage-Weighted Regression: Simple and Scalable Off-Policy Reinforcement Learning
• ADaPT: As-Needed Decomposition and Planning with Language Models
• ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs
• The Last Harness You'll Ever Build
• Reflexion: Language Agents with Verbal Reinforcement Learning
• AutoPDL: Automatic Prompt Optimization for LLM Agents
• ScoreFlow: Mastering LLM Agent Workflows via Score-based Preference Optimization
• AgentGym: Evolving Large Language Model-based Agents across Diverse Environments
• SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering
• $\tau$-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks