🧪AI-Scientist-v2
The AI Scientist-v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Search
⭐ Popular: 7k stars, gaining about 32 a week
View on GitHub ↗repo profile
momentum
durability
since we covered it
why it's a big deal
- It automates a full research loop from idea generation through experiments to a written manuscript, targeting machine learning researchers who want to offload exploratory studies.
- The project reports producing the first workshop paper written entirely by AI and accepted through peer review, which sets a concrete reference point for what these systems can currently do.
- Unlike v1 it drops human-authored templates, so it aims to generalize across machine learning domains rather than run only pre-scaffolded experiments.
under the hood
- Written in Python and split into two stages; perform_ideation_temp_free.py generates and novelty-checks ideas via Semantic Scholar, then launch_scientist_bfts.py runs experiments and writes the paper.
- Experiments are driven by a best-first tree search guided by an experiment manager agent, with configurable num_workers, steps, and max_debug_depth.
- It mixes several models by role, using Claude for experimentation, GPT-4o for citations, and o1-preview or o3-mini for writing, and runs on PyTorch with CUDA alongside LaTeX tooling like poppler and chktex.
Radar summary, generated from the project's public sources
star history
- PR#31 5k 2026-04-08
- now 7k + 2k since first covered
curve is sampled from GitHub's star history, plus our own daily readings since we covered it; the dashed stretch is before we first covered it, the solid line since. figures at coverage are the numbers we printed then (approx.), current count is live.
understory
Better known than its recent output, coasting a little on attention.
- output, commits & releases
- clout, star velocity
output = commits & releases; clout = star velocity, both 0 to 100 monthly indices; the gap where output runs above clout is the understory. The understory →
covered in
-
End-to-end agentic system for autonomous scientific paper generation via tree search
similar projects
compare these →- ☤ hermes-agent
33× the stars
The agent that grows with you
231k ACTIVE - 🖼️ Langflow
22× the stars
Langflow is a powerful tool for building and deploying AI-powered agents and workflows.
153k ACTIVE - 📊 TradingAgents
14× the stars
TradingAgents: Multi-Agents LLM Financial Trading Framework
98k ACTIVE
comments
Sign in with GitHub to add your blip on AI-Scientist-v2.