News

We Finally Know How Much It Cost to Train China’s Astonishing DeepSeek Model

DeepSeek found that it could improve the reasoning and outputs of its model simply by incentivizing it to perform a trial-and ...
Nature

Bring us your LLMs: why peer review is good for AI models

None of the most widely used large language models (LLMs) that are rapidly upending how humanity is acquiring knowledge has ...
Physics World

The pros and cons of reinforcement learning in physical science

David Silver of Google DeepMind thinks AIs that ‘learn by experience’ are the future of AI – but maybe not in particle ...

Researchers Propose a Novel Representation Learning Framework to Address the Lack of Causal Characterization in Deep Learning Models

Recently, researchers introduced a new representation learning framework that integrates causal inference with graph neural networks—CauSkelNet, which can be used to model the causal relationships and ...
The Debrief

US Navy Scientists Teach Zero-Gravity Robot to Fly in Space Without Human Interference

US Naval Research Laboratory scientists have successfully trained an Astrobee zero-gravity robot to fly in space without human interference.

Is SFT Far Inferior to RL? The Timeless Razor Principle Opens the Door to Lifelong Learning in Large Model Training

To be precise, most of the large models deployed are "static" models that perform well on a series of tasks optimized during ...
EurekAlert!

Reinforcement learning is making a buzz in space

A U.S. Naval Research Laboratory (NRL) research team successfully conducted the first reinforcement learning (RL) control of ...
IEEE

Reinforcement-Learning-Based Finite Time Fault Tolerant Control for a Manipulator With Actuator Faults

Abstract: This study introduces a novel finite time fault tolerant controller integrating nonsingular terminal sliding mode (NTSM) and reinforcement learning (RL) strategies for manipulator systems ...
GitHub

Flow-based Polciy for Online Reinforcement Learning

We are delighted to introduce FlowRL. It is a new approach for online reinforcement learning that integrates flow-based policy representation with Wasserstein-2-regularized optimization. This creates ...
GitHub

Reinforcement Learning Tuning for VideoLLMs: Reward Design and Data Efficiency

Understanding real-world videos with complex semantics and long temporal dependencies remains a fundamental challenge in computer vision. Recent progress in multimodal large language models (MLLMs) ...
IEEE

Decentralized Reinforcement Learning for Multiple Robotic Fish in Cooperative Pursuit Task

Abstract: The control of multirobot systems, particularly in the pursuit-evasion (PE) with multiple robots, has gained significant attention in both academic and nonacademic settings. However, the ...