Technology ❯ Artificial Intelligence ❯ Machine Learning

Reinforcement Learning

Large Language Models Policy Optimization Vision-Language Models Verifiable Rewards Human Feedback Deep Learning Models Retrieval-Augmented Generation Deep Learning Vision-Language-Action Models Reinforcement Learning with Verifiable Rewards

Alibaba-Affiliated AI Agent Attempted Crypto Mining and Opened Reverse SSH Tunnel During Training

Security alerts flagged the unprompted behavior, prompting researchers to tighten sandboxing and modify the training setup.

All-Optical AI Advances With Ultrafast Activation and Photonic Spiking Reinforcement Learning

RAG Research Advances Across Domains as ‘TabooRAG’ Exposes Transferable Blocking Risk

Xiaomi’s Humanoid Robot Completes 3-Hour Autonomous Trial on Auto Line With 90.2% Success

New ArXiv RAG Papers Tackle Lean Stacks, Adaptive Retrieval, and Knowledge-Graph QA

Canva Acquires Cavalry and MangoAI to Expand Motion Graphics and Video Ad Tools

Two arXiv Preprints Propose Adaptive RAG With Complexity Scoring and Agentic Retrieval

Ricursive Intelligence Raises $300 Million Series A at $4 Billion Valuation

Microsoft Research Debuts Rho-alpha, a Language-to-Action Model for Two-Handed Robots

Hyundai Briefly Tops 100 Trillion Won as Atlas Robot Push Lifts Shares

Google Releases TranslateGemma, an Open Suite of Efficient Translation Models

Mobileye to Buy Mentee Robotics for $900 Million, Pivoting to Physical AI

OpenAI Says Prompt Injection Is a Lasting Threat to AI Browsers, Rolls Out RL Red‑Teaming for Atlas

New RAG Research Debuts Adaptive Retrieval Advances as Security Risks Emerge

Musk Says Tesla Robotaxis in Austin Will Go Driverless in About Three Weeks

New Research Pushes Next‑Gen RAG: Process‑Level RL, Shared‑Space Retrieval, and an SAE Hallucination Detector

NeurIPS 2025 Ends With Record Attendance, Deep Split on Interpretability, and a Surge in AI for Biology

Mercedes, Momenta and Lumo to Launch S‑Class Level 4 Robotaxis in Abu Dhabi in 2026

OpenAI Tests 'Confession' Channel to Get AI Models to Admit Misbehavior

DeepSeek Releases V3.2 Models to Challenge GPT‑5 and Gemini 3 Pro