AI-Scientist-V2

Visit Tool

AI-Scientist-v2 is an AI tool for automated scientific discovery that uses agentic tree search to generate hypotheses, run experiments, analyze data, and write scientific manuscripts. It aims to transform how scientific discoveries are made.

Claim this tool

2Views

At a glance

Pricing

Open Source

Free tier

Yes

API

Skill level

Technical

About

What is AI-Scientist-v2?

The AI Scientist-v2 is a generalized end-to-end agentic system designed for fully autonomous scientific research, capable of workshop-level automated scientific discovery. Unlike its predecessor, v2 removes reliance on human-authored templates and generalizes across Machine Learning (ML) domains. It employs a progressive agentic tree search guided by an experiment manager agent to autonomously generate hypotheses, run experiments, analyze data, and write scientific manuscripts. The system supports OpenAI, Gemini, and Claude models via AWS Bedrock, and can optionally use Semantic Scholar API for literature search. Users can generate research ideas based on a topic description and then launch the main pipeline to conduct experiments and produce a paper draft.

Best used for

Ideal for professors and data scientists who need to accelerate scientific discovery, automate experimental pipelines, and generate research papers. Especially valuable for exploring open-ended scientific questions and reducing the manual effort in the research process.

Common actions

automate research

generate hypotheses

run experiments

analyze data

write scientific papers

github copilot"AI Agents"face swappingcollaborationdeepfakeopen-sourceautomated workflowlow-code/no-codeworkflows

Capabilities

Key features

Automated scientific discovery
Agentic tree search
Hypothesis generation
Experiment execution
Data analysis
Scientific manuscript writing
LLM-driven ideation

Target Audience

professordata scientistdeveloper

Integrations

Not yet documented

Pricing & Plans

Open Source

Free

FAQs

What are the system requirements for running AI-Scientist-v2?

The code is designed to run on Linux with NVIDIA GPUs using CUDA and PyTorch. Installation requires setting up a conda environment, installing PyTorch with CUDA support, and installing PDF/LaTeX tools like Poppler and ChkTeX, along with Python package requirements.

What is the estimated cost per experiment run?

The ideation step typically costs a few dollars. For the main experiment pipeline, using Claude 3.5 Sonnet costs around $15–$20 per run, with the writing phase adding approximately $5. Costs can vary based on the LLM used and the complexity of the research idea.

How can I ensure the safety of running LLM-written code?

The codebase executes LLM-written code, which carries risks like dangerous packages or unintended processes. It is strongly recommended to run AI-Scientist-v2 within a controlled sandbox environment, such as a Docker container, to mitigate these risks.

How does AI-Scientist-v2 differ from its predecessor, AI Scientist-v1?

AI-Scientist-v2 removes reliance on human-authored templates, generalizes across ML domains, and uses a progressive agentic tree search. While v1 excels with clear objectives and templates, v2 takes a broader, more exploratory approach, designed for open-ended scientific exploration.

Trending

Subcategories trending in Research & Education

Academic Research Study Assistants Knowledge Management Course Creation Summarization Language Learning

Trending

Explore

Browse AI tools by category

Content & Design Productivity & Business Coding & Development AI Agents & Automation Research & Education Wellness & Lifestyle Career Development Marketing & Growth Data & Analytics Customer Support & CX Finance E-commerce