AMD Researchers Introduce Agent Laboratory: An Autonomous LLM-based Framework Capable of Completing the Entire Research Process
autonomous-researchllm-agentsscientific-automationresearch-pipelineamd
Abstraction: Autonomous LLM pipeline completing literature review, experiments, and paper writing
Key points:
- Agent Laboratory (AMD + Johns Hopkins) uses specialized LLM agents — "PhD" for literature review, "ML Engineer" for experimentation, "Professor" for writing — to automate the full research lifecycle
- Reduces research costs by up to 84% and shortens project timelines via task automation
- Using o1-preview backend achieves 95.7% task success rate; GPT-4o backend completes projects for as little as $2.33
- mle-solver module outperforms human baselines on MLE-Bench and earns competition medals autonomously
- Co-pilot mode integrating user feedback produces the most impactful research outputs
- Paper: arxiv 2501.04227; code at github.com/SamuelSchmidgall/AgentLaboratory
Connections: Amd · AI Agents · Large Language Models