The AI Scientist: Automating Research Plan Creation

A new framework uses artificial intelligence to autonomously generate detailed research plans, bypassing the need for human-created training data.

A new framework uses artificial intelligence to autonomously generate detailed research plans, bypassing the need for human-created training data.

A new Python library streamlines the process of creating stunning robotic visualizations and animations directly within the popular open-source 3D creation suite.
Researchers have released an open-source evaluation toolkit designed to rigorously assess the scientific intelligence of artificial intelligence models, uncovering critical limitations in their ability to reason and problem-solve within complex scientific domains.
A new perspective argues that building robots capable of understanding human intentions requires a more rigorous evaluation of how they explain their actions.
A new analysis reveals the key factors driving research partnerships in the fight against cancer, leveraging machine learning to understand collaborative patterns.

This review explores how recent advances in artificial intelligence are transforming robotic manipulation, enabling robots to perform increasingly complex tasks.
A comprehensive analysis reveals how artificial intelligence is being applied – and where it still needs to be explored – within the world of lean startup methodologies.

Researchers are drawing inspiration from insect swarms to create more resilient and adaptable collective robot motion systems.
![An integrated framework for agentic physical AI demonstrates that scaling foundation models from 1K to 100K nuclear reactor scenarios induces qualitative phase transitions-increasing precision from 26.2% to 92%, collapsing variance by 500×, and compressing policy entropy from 1.38 to 0.89 nats-while a two-phase curriculum leveraging CPT and LoRA stabilizes agentic policies by separating domain structure from task specialization and concentrating 76% of actions on single strategies despite limited training frequency, ultimately achieving closed-loop control within specified tolerances in a physics-constrained environment [latex]\mathcal{M}\_{\text{feas}}[/latex].](https://arxiv.org/html/2512.23292v1/x1.png)
A new approach to artificial intelligence demonstrates reliable power control of a nuclear reactor through data-driven learning and closed-loop simulation.

New research maps the design landscape for AI agents that collaborate with humans in visual analytics, offering a systematic approach to building more effective data exploration tools.