Alibaba's Metis agent cuts redundant AI tool calls from 98% to 2% — and gets more accurate doing it

Alibaba's HDPO framework trains AI agents to skip unnecessary tool calls, cutting redundant invocations from 98% to 2% while ...

Why OpenAI's 'goblin' problem matters — and how you can release the goblins on your own

If OpenAI can accidentally train its flagship model to obsess over goblins, what other more subtle and potentially harmful ...
Hosted on MSN

How AI Is Rewriting the Rules of Excel and Business Computing in 2026

If you are a university student preparing for a career in finance or data analytics, the skills you are learning in your business computing and Excel courses are undergoing a radical transformation.
Futurism

Sam Altman’s Coworkers Say He Can Barely Code and Misunderstands Basic Machine Learning Concepts

Sam Altman, OpenAI’s CEO and the public face of ChatGPT, has carved out an image for himself as one of the preeminent AI whisperers of our age, whose influence supposedly extends to the White House on ...
techxplore

New memristor design uses built-in oxygen gradient to bring stability to reinforcement learning

In a recent study published in Nature Communications, researchers created a memristor that uses a built-in oxygen gradient to produce slow, stable conductance changes, enabling a reinforcement ...
IEEE

RLCoder: Reinforcement Learning for Repository-Level Code Completion

Abstract: Repository-level code completion aims to generate code for unfinished code snippets within the context of a specified repository. Existing approaches mainly rely on retrievalaugmented ...
IEEE

A Reinforcement Learning Environment for Automatic Code Optimization in the MLIR Compiler

Abstract: Code optimization is a crucial task that aims to enhance code performance. However, this process is often tedious and complex, highlighting the necessity for automatic code optimization ...
Forbes

Leadership Amid Uncertainty: CEOs Can Learn Effective Decision Making From Reinforcement Learning

Leaders, whether in boardrooms or garages, constantly face an unchanging force: uncertainty. For a CEO, making a good decision always involves factoring in as much data as possible, and then trusting ...
Microsoft

Argos: Multimodal reinforcement learning with agentic verifier for AI agents

Over the past few years, AI systems have become much better at discerning images, generating language, and performing tasks within physical and virtual environments. Yet they still fail in ways that ...
Hosted on MSN

Watch an AI learn to balance a stick — reinforcement learning in action

Watch an AI agent learn how to balance a stick—completely from scratch—using reinforcement learning! This project walks you through how an algorithm interacts with an environment, learns through trial ...
Microsoft

Agent Lightning: Adding reinforcement learning to AI agents without code rewrites

AI agents are reshaping software development, from writing code to carrying out complex instructions. Yet LLM-based agents are prone to errors and often perform poorly on complicated, multi-step tasks ...