Composing Task-specific Agent Harnesses at Test Time with Reusable Primitives
cs.AI
Submitted: 2026-09-30
Updated: 2026-09-30
Sources
- Evaluating Large Language Models Trained on Code
- Automated Design of Agentic Systems
- MemoHarness: Agent Harnesses That Learn from Experience
- SWE-bench: Can Language Models Resolve Real-World GitHub Issues?
- KV-PRM: Efficient Process Reward Modeling via KV-Cache Transfer for Multi-Agent Test-Time Scaling
- Recursive Harness Self-Improvement
- Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces
- TTHE: Test-Time Harness Evolution
- Code as Agent Harness
- HARBOR: Automated Harness Optimization
- Reflexion: Language Agents with Verbal Reinforcement Learning
- OpenHands: An Open Platform for AI Software Developers as Generalist Agents
- Better Harnesses, Smaller Models: Building 90% Cheaper Agents via Automated Harness Adaptation
- SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering
- JIT-Agent: Scaling Harness Intelligence via Just-in-Time Harness Evolution
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