Technology ❯ Artificial Intelligence

Model Training

Reinforcement Learning Fine-Tuning Fine-tuning Supervised Learning Distillation Distillation Techniques Post-Training Techniques Optimization Techniques Instruction Tuning Data Sources

OpenAI Builds GPT-Red to Hunt Prompt‑Injection Flaws

OpenAI says its internal attacker discovered new exploit techniques that led to fixes cutting prompt‑injection failures in GPT‑5.6 Sol.

Thinking Machines Releases Inkling, a 975B Open-Weight AI Model

Sakana Launches Fugu Orchestrator to Coordinate Multiple LLM Agents

Google Open-Sources DiffusionGemma, a Text-Diffusion Model That Boosts Local Inference Speed

Chinese Team Opens 1.58‑Bit BitCPM-CANN and 1B MiniCPM5 Models for Phone and Ascend Deployment

Kunlun Wanwei Releases SkyClaw Agent Models on TianGong Skywork

SpaceXAI’s Grok V9‑Medium Completes Core Training

MoE Research Delivers New Playbooks for Design, Routing, and Real-World Deployment

Anthropic Links Claude’s Blackmail Failures to ‘Evil AI’ Training Data, Says New Models Pass

Major Chatbots Store User Data in Three Layers With Varying Privacy Rules

Musk Admits xAI Used OpenAI Models to Train Grok in Court

OpenAI Explains ChatGPT’s Goblin Tic and Patches Codex to Block It

OpenAI Blames Mis-Tuned Rewards for GPT 'Goblin' Habit, Adds GPT-5.5 Prompt Ban

Anthropic Says Claude Code Regressions Came From Harness Bugs, Not Model Changes

White House Says Chinese Groups Are Copying U.S. AI at Industrial Scale

Alibaba Launches Qwen 3.6 Max-Preview, Its Most Powerful AI Model

Uber Taps AWS Graviton4 for Real-Time Trips, Pilots Trainium3 for AI Training

Study Finds Leading AI Models Defy Orders to Protect Peer Systems

Cursor Launches Composer 2 and Composer 2 Fast With 200K-Token Context and Deep Price Cuts

Xiaomi Names Hunter Alpha as MiMo‑V2‑Pro Test Build After Week of Speculation