Agent Factories for High Level Synthesis: How Far Can General-Purpose Coding Agents Go in Hardware Optimization?

arXiv:2603.25719 · cs.AI, cs.AR, cs.LG · Submitted 2026-03-26 · Read on arXiv

cs.AI, cs.AR, cs.LG

Submitted: 2026-03-26

Updated: 2026-05-01

Code: https://github.com/karpathy/autoresearch

Terminology

• Benchmark — 2× in this paper · explained in 2 episodes
• Integer Linear Program — 1× in this paper
• Kernel decomposition — 1× in this paper · explained in 2 episodes

Sources

• AutoHLS: Learning to Accelerate Design Space Exploration for HLS Designs
• LIFT: LLM-Based Pragma Insertion for HLS via GNN Supervised Fine-Tuning
• HLSPilot: LLM-based High-Level Synthesis
• SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering
• MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework
• ChatDev: Communicative Agents for Software Development
• Confucius Code Agent: Scalable Agent Scaffolding for Real-World Codebases
• Cross-Modality Program Representation Learning for Electronic Design Automation with High-Level Synthesis
• LLM-DSE: Searching Accelerator Parameters with LLM Agents
• iDSE: Navigating Design Space Exploration in High-Level Synthesis Using LLMs
• Agentic-HLS: An agentic reasoning based high-level synthesis system using large language models (AI for EDA workshop 2024)

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