#Research

Offline Learning Without Sacrifices: How Conservative Q-Learning Saves Budgets and Lives

CVPR 2026: Visual agents learn to survive under hackers' fire

Budget Thriller in Washington: AI Industry Finally Catches Its Breath

Research·Bloomberg Tech

Butterfly Effect Under Threat: How Google and Nvidia Are Trying to Tame Weather Chaos

Meteorology has always been a science of deep humility. We are accustomed to the fact that tomorrow's forecast is useful information, while

01Vector Magic: 7 Ways to Maximize LLM EmbeddingsFeb 3, 2026·2 min 02RebuttalAgent: AI Learned to 'Read Reviewers' Minds' (and Save Your Papers)Feb 3, 2026·2 min 03Google DeepMind Takes Silver: AI Solves Olympiad Problems, But at What Cost?Feb 3, 2026·2 min 04Moltbook: why people dream of becoming bots in their own social networkFeb 3, 2026·2 min 05Neural Networks at the Collider: AI Searching for Physics We Didn't OrderFeb 3, 2026·3 min

Top themes

Battle for Style: Which AI Will Rewrite Your Text Better Than Humans

Remember those times when text rewriting meant simply rearranging words using questionable synonym services? Those days are officially dead.

01Truth Crisis in the Era of LLMs: Why Your AI Assistant Is a Pathological LiarFeb 3, 2026·2 min 02Time Series: Five Neural Networks That Forecast the Future Better Than AnalystsFeb 3, 2026·2 min 03Yao Shunyue at Tencent: why the leading AI research star started with the basicsFeb 3, 2026·2 min

AI agents: convenient helper or open door for hackers?

Let's be honest: we all waited for the moment when AI would stop being just a chatty chatbot and start *doing* things. Booking tickets, debu

Research·36Kr (36氪)

GE Aerospace and Singapore gambit: $300 million for AI to inspect your engines

Imagine you're flying over the ocean at ten thousand meters altitude, and the only thing between you and the water is a huge piece of metal

AI-Agents Off the Leash: Why It's Time to Forget About Hard Routing

Imagine you're building a complex maze of logical conditions where each turn depends on a language model's response.

Breathe Deeply: Why a Russian Neural Network Needs to Monitor Your Chest

AI-Agents 2030: When Your Software Stops Asking Permission

Russian Voice from a Box: Why Open Source TTS No Longer Sounds Like a 90s Robot

EMNLP 2025: Why neural networks now check themselves (and why it's hard)

Research·Guardian

International Security Report: AI learned to hack and befriend

Robot Bolt: China taught humanoids to run faster than Olympic athletes

Armor for Neural Networks: Why Your LLM Needs More Than One Security Filter

304 Chinese LLMs: Why Among Hundreds of Neural Networks They Haven't Found a King

Research·Jiqizhixin (机器之心)

SoulX-FlashTalk: Chinese Soul App Makes Digital Doubles Talk Without Delay

Remember those days when digital avatars looked like poorly dubbed films from the eighties? Lips had a life of their own, and facial express

Research·36Kr (36氪)

Zhipu GLM-OCR: How Chinese Tech Giant Taught Micro-Model to See Everything

The artificial intelligence industry has long resembled a bodybuilders' off-season competition: each new announcement was accompanied by bra