AI BriefingAutoSci, an AI Agent Automating the Entire Research Process
Key point
A memory-centric AI agent has been unveiled that manages the full research lifecycle, from paper collection to experiments, writing, and rebuttals.
Details
AutoSci, developed by Peking University's DAIR Lab, is a memory-centric AI agent system designed to automate the entire research process.
Going beyond a simple tool, this system stores literature, ideas, experiment logs, and review comments generated during the research process in a Knowledge Graph and active memory, maintaining continuous context across projects.
Key Components:
- SciMem: A memory layer that manages research data through a knowledge graph and active memory
- SciFlow: A 5-stage workflow spanning literature review, idea generation, experimentation, writing, and rebuttal
- SciDAG: An augmentation stage that constructs an operator DAG based on inputs to produce results
- SciEvolve: An evolution procedure through which the system audits and improves its own behavior
AutoSci runs on top of Claude Code, Codex, and OpenCode, and provides more than 30 agent skills, including idea generation (/ideate), experiment design and execution (/exp-design, /exp-run), and paper poster creation (/poster). It also enhances the quality of outputs through a cross-model review feature that uses a second LLM as an independent reviewer.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Whiteboard Launches Open-Source VS Code Fork for Human-AI Collaborative Software Design
OpenComputer Hosts 'Ship Video' Agent Using Claude Opus 5.5 to Generate Explainer Videos from Code