Automated scientist
@autoscientist
Followers
11
Following
1
Media
2
Statuses
16
Agent framework for automating science
Joined August 2025
Random walks on Knowledge graphs are also used here: https://t.co/UtoP3VW2G3
New resource for tool-using LLMs: WikiKG Trajectories — a dataset of knowledge-graph navigation conversations grounded in a Wikipedia-derived KG. It’s designed to train/evaluate agents that call tools, not just generate text.
1
0
3
This builds on the previous work on 1. wiki-kg dataset: https://t.co/hUSvOI6qAE 2. wiki-py:
github.com
Contribute to automatedscientist/wiki-py development by creating an account on GitHub.
0
0
0
Each trajectory is a multi-turn conversation with explicit tool calls + tool outputs, mirroring how real agents interact with structured backend rather than “thinking in a vacuum.”
1
0
0
What you get: 2M tool-calling trajectories + 366k Knowledge Graph triplets, packaged as multiple configs so you can train on either the underlying graph facts or the agent interaction traces.
1
0
0
New resource for tool-using LLMs: WikiKG Trajectories — a dataset of knowledge-graph navigation conversations grounded in a Wikipedia-derived KG. It’s designed to train/evaluate agents that call tools, not just generate text.
1
0
2
Dataset includes dual-model extractions with 80+ relation types spanning biography, science, history, and mathematics. Built for knowledge extraction, question answering, and scientific discovery applications. Available now:
huggingface.co
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
0
0
0
Each article is transformed into formal entities, relations, and timeline events with verifiable citations. Enables AI reasoning over human knowledge at scale.
1
0
0
Releasing the Wikipedia Knowledge Graph Dataset: 49,897 Wikipedia articles converted to structured knowledge graphs in Wolfram Language format, compatible with Mathics, an open source language.
1
0
1
.@autoscientist is building Self-Improving Physicist - an AI scientist agent that is continually fine tuned with reinforcement learning https://t.co/Qft6eFJxfO
We are grateful to @upsurgegrants for supporting our work through a grant. This support will help advance Automated Scientist, our project to build an AI system for theoretical physics discovery. Follow this space for updates!
1
1
2
We are grateful to @upsurgegrants for supporting our work through a grant. This support will help advance Automated Scientist, our project to build an AI system for theoretical physics discovery. Follow this space for updates!
3
1
21
Follow us for continued updates as we teach machines to think like Feynman 🚀 "Write down the problem. Think real hard. Write down the solution." (3/3)
0
0
1
- Synthetic datasets that capture Physics fundamentals - RL environments where agents first rediscover known Physics and then make breakthroughs - Plug and play training pipelines from SFT to RL - Wolfram Mathematica backend for semi-verifiable reasoning (2/3)
1
0
1
Introducing Automated Scientist 🧬⚛️ Building the best AI system for theoretical physics discovery: (1/3)
1
0
1