Breaking Papers
The most important scientific papers, decoded. 309 papers analyzed from arXiv and beyond.
The most important scientific papers, decoded. 309 papers analyzed from arXiv and beyond.
Silicon quantum computing keeps promising that manufacturability will save it.
Researchers from Zhejiang University, Westlake, and Beijing AI propose VAMPO, a reinforcement learning approach to fix the 'looks right, fails anyway' problem in diffusion-based video action models.
The attack does not look like an attack.
Insilico Medicine, the Hong Kong-based AI drug discovery company that went public on the Hong Kong Exchange in December 2025, has launched PandaClaw — an agentic AI system purpose-built for biologists working on therapeutic discovery.
The problem with robot navigation is that it's never really portable.
The safety problem in soft robotics has always had two layers.
Every soft robot ever built with integrated sensing has come with a hidden cost: before it could feel the world, someone had to teach it what its own body felt like.
A new framework called PhyGile tackles one of humanoid robotics' sneakiest problems: motions that look perfect on a human and faceplant on a machine. The robots keep falling.
A team from George Washington University and Johns Hopkins has published a machine unlearning method designed specifically for clinical language models — one that nearly eliminates a target class from a deployed model while modifying less than 0.2% of its parameters.
A team spanning UBC, Oxford, and Meta has published a paper extending the Darwin Godel Machine into a system where not just the task-solving code, but the improvement mechanism itself, can be rewritten at runtime.
Running a mixture-of-experts model in production has a quiet bottleneck that benchmark papers rarely discuss: the CPU-GPU transfer.
Drone swarms have a communication problem, and the solution may be light. Researchers at Worcester Polytechnic Institute (WPI), a technical university in Worcester, Massachusetts, and MIT Lincoln Laboratory, a federally funded research and development center operated by MIT for the U.S.
Proving that a robot path-planning algorithm is within a known factor of optimal is not the kind of work you hand to an intern.
Oxford's latest ZX-calculus paper is not a new quantum decoder, and that is exactly why it matters.
Quantum computing does not get less noisy because a paper learns a longer adjective.
Soft robot fingers are hard to simulate.
The robot arm freezes mid-task.
