MACHINE LEARNING

→ View original post on X — @ylecun

French pioneers: Model predictive control and system identification history

By

@ylecun

–

02 February 2025 21h03

Hey dudebot, you wanna play that game?
The first publication about model predictive control with system identification (learned world model) is from a bunch of French dudes in 1978. You and I were still in high school.
They proposed a method called IDCOM, which I learned

→ View original post on X — @testingcatalog

Qwen Narrows Gap Between Plus and Max

By

@testingcatalog

–

02 February 2025 20h27

Qwen has updated its Qwen2.5-Plus model, reducing the performance gap between Plus and Max—the most advanced Qwen model available.

→ View original post on X — @ronald_vanloon

Measuring Generative AI ROI: A Practical Guide

By

@ronald_vanloon

–

02 February 2025 18h53

Unlocking generative #AI’s true value: a guide to measuring ROI
by @jathomason @VentureBeat Read more: https://
buff.ly/3ADNoKp #MachineLearning #ArtificialIntelligence #ML #MI #Tech #Technology cc: @meisshaily @ylecun @ogrisel

→ View original post on X — @hwchase17

Continual Learning in MLOps and Agent Building Platforms

By

@hwchase17

–

02 February 2025 18h39

The holy grail of MLOps was always continual learning The same concepts are popping up in agent building ("agent ops"?) Learn more about it this Tuesday (in SF only!) when Decagon talks about their agent platform https://
lu.ma/w7y0bqwr

→ View original post on X — @ronald_vanloon

DEEPRobotics Showcases Advanced Capabilities of Lynx Robot

By

@ronald_vanloon

–

02 February 2025 17h58

DEEPRobotics shows off the Lynx robot's capabilities
by @IntEngineering #Robotics #AI #ArtificialIntelligence #MachineLearning #MI #Tech #Technology

cc: @kirkdborne @miketamir @ogrisel pic.twitter.com/9WOS6oRqTd
— Ronald van Loon (@Ronald_vanLoon) 2 février 2025

DEEPRobotics shows off the Lynx robot's capabilities by @IntEngineering #Robotics #AI #ArtificialIntelligence #MachineLearning #MI #Tech #Technology cc: @kirkdborne @miketamir @ogrisel

TensorLLM Framework Achieves 250x MHA Weight Compression

By

–

02 February 2025 16h51

9). TensorLLM Proposes a framework that performs MHA compression through a multi-head tensorisation process and the Tucker decomposition. Achieves a compression rate of up to ∼ 250x in the MHA weights…

Docling: Open-Source Document Parsing Toolkit

By

–

02 February 2025 16h51

7). Docling Docling is an open-source toolkit that can parse several types of popular document formats into a unified, richly structured representation. https://
arxiv.org/abs/2501.17887

Improving RAG Through Multi-Agent Reinforcement Learning

By

–

02 February 2025 16h51

8). Improving RAG through Multi-Agent RL It models RAG components like query rewriting, document selection, and answer generation as RL agents working together toward generating accurate answers.

Diverse Preference Optimization: Novel Training Method for Language Models

By

–

02 February 2025 16h51

5). Diverse Preference Optimization A novel training method that aims to address the lack of diversity in language model outputs while maintaining response quality.