Open Reasoning Data

General Reasoning | Where Machines Get Reward is an AI research initiative focused on building safe and capable reasoning models through reinforcement learning. Based in London, the team aims to push the frontier of machine reasoning by developing scalable, open approaches to intelligent behavior and governance.

Overview

General Reasoning emphasizes advancing reinforcement learning-driven reasoning to produce robust and safe AI systems. The research agenda includes developing novel reasoning architectures, scalable training workflows, and practical deployments that align with human intentions. Key items in the initiative include open data, community collaboration, and transparent research practices.

Publications and Artifacts

• GeneralThought-430K: a large-scale dataset (announced 14 March 2025) to support reasoning-focused research.
• Blog: covering insights on scaling reinforcement learning for reasoning tasks.
• Open Data Hub (launched 21 February 2025): a platform for sharing datasets, experiments, and research outputs to foster open science.

Collaboration and Team

General Reasoning is seeking to build its initial team for the first phase of the company. The team is based in London but welcomes remote collaboration and contributions from diverse researchers and engineers. Interested candidates can submit their resume to join the team.

Policies and Governance

• Terms of Service
• Privacy Policy
• Content Policy

How to Engage

• Explore the blog for research updates and thought leadership.
• Access the Open Data Hub to view or contribute datasets and artifacts.
• Review current openings and apply with a resume if you are interested in joining the team.

Core Features

• Reinforcement learning-driven reasoning models for improved AI safety and capability
• Open Thought datasets (e.g., GeneralThought-430K) to advance research in reasoning
• Open Data Hub for sharing datasets, experiments, and research outputs
• Public-facing blog with insights on scaling RL for reasoning
• London-based organization open to global collaboration
• Clear governance and policy documentation (Terms, Privacy, Content Policy)