Key AI papers, new benchmarks and methods, and what AI is discovering in biology, medicine, materials and other sciences.
Researchers just introduced VideoJAM, a framework designed to tackle AI video's persistent struggle with real-world physics and motion coherence. By training models on joint appearance-motion representations and using internal motion cues to steer generation during inference, it plugs into existing architectures and outperforms proprietary baselines like Sora on complex motion.
Researchers just introduced MatAnyone, a memory-based video matting framework tailored for complex scenes. By simply providing a segmentation mask on the first frame, the model leverages region-adaptive memory propagation to maintain semantic tracking while preserving delicate alpha boundaries. It aims to make Hollywood-level object separation accessible with just a few clicks.
Yale School of Medicine researchers just built an AI tool that predicts future heart failure risk using standard 12-lead ECG images. The model was successfully validated across patient populations in the US, UK, and Brazil, giving doctors a low-cost screening method well before overt symptoms show up.
Microsoft Research just published and open-sourced MatterGen, a generative diffusion model that designs novel physical materials directly from property prompts instead of screening databases. Working with SIAT researchers, the team successfully synthesized a completely new material (TaCr2O6) in the lab with under 20% error against specifications. Why it matters: It shifts material discovery from brute-force search to generative design for batteries and solar cells.
Researchers just introduced a new framework that scales diffusion models at inference time beyond traditional denoising steps. By pairing automated verifiers with search algorithms to hunt for better noise candidates, dumping more compute into generation now reliably translates into higher image quality.
Keras creator François Chollet just teamed up with Zapier co-founder Mike Knoop to launch Ndea, a new research lab betting on program synthesis over raw model scaling. The team aims to combine deep learning with automated code generation so AI can adapt and reason like humans. Why it matters: Instead of throwing more compute at the same LLM architectures, Chollet is building an alternative path toward true intelligence.
David Baker’s lab just used deep learning to design brand-new proteins that neutralize lethal snake venom. In mouse tests published in Nature, the computationally generated molecules yielded up to a 100% survival rate against deadly cobra neurotoxins. Traditional antivenoms rely on costly animal plasma, but these compact AI proteins can be brewed cheaply in microbes. Why it matters: While it can't neutralize full venom cocktails yet, it paves the way for accessible, life-saving antidotes in developing nations.
Sakana AI just unveiled Transformer², an adaptive system that adjusts LLM weights at test time to tackle different tasks like a living brain. Using Singular Value Finetuning and RL, it beats LoRA across math and coding benchmarks while requiring far fewer parameters.
Netflix and Eyeline researchers just introduced "Go with the Flow," a new method that steers motion in video diffusion models without changing model architectures. By warping latent noise using optical flow fields in real time, it enables precise camera moves and object control while keeping frame quality crisp.
Google DeepMind is building a new team dedicated to AI that simulates the physical world, led by former OpenAI Sora co-lead Tim Brooks. The group plans to build on Gemini, Veo, and Genie to advance visual reasoning, robotics planning, and real-time generation, viewing world models as essential on the path to AGI.
Meta researchers just detailed how generative retrieval can predict user intent instead of relying on traditional database searches. By introducing frameworks called LIGER and Mender, the system treats recommendations as sequence predictions, slashing storage costs while using LLMs to capture subtle user preferences. Why it matters: It shows generative models could soon replace the classic embedding systems powering the internet's biggest feeds.
Researchers just introduced Titans, a new model architecture that pairs attention-based short-term memory with a neural long-term memory module that learns at test time. It scales past a 2 million token context window while keeping training and inference fast, outperforming standard Transformers on needle-in-haystack tasks.
OpenAI cofounder Ilya Sutskever just made a rare public appearance at NeurIPS, declaring that AI has officially achieved "peak data." Comparing human web data to finite fossil fuels, he warned traditional pre-training is nearing its end. Future breakthroughs, he argued, will rely on truly reasoning and agentic models rather than mass scraping.
Anthropic just shared new research on "Best-of-N Jailbreaking," uncovering a simple, general-purpose method that bypasses safety guardrails across frontier AI models. The attack works across text, vision, and audio, underscoring persistent blind spots in multimodal safety defenses.
Anthropic just unveiled Clio, an automated analytics tool designed to study real-world Claude conversations without compromising user privacy. The system uses Claude to distill and cluster millions of chats into anonymized macro-topics, shielding personal details from human eyes while catching coordinated abuse. Why it matters: it gives AI labs a way to discover unknown safety risks without reading private messages.
IBM just unveiled a breakthrough in co-packaged optics (CPO), introducing the first polymer optical waveguide prototype to replace copper wires inside data centers. The tech can slash standard LLM training times from three months to three weeks while cutting power consumption fivefold. It has also passed all key manufacturing stress tests.
Sakana AI just introduced Neural Attention Memory Models (NAMMs), a biologically inspired memory system that allows transformers to selectively discard redundant tokens. Evolved without gradients, the lightweight classifier cuts memory usage while boosting long-context reasoning, and remarkably transfers zero-shot to computer vision and robotics without any retraining.
NVIDIA Research and the University of Maryland just introduced QUEEN, an AI model that enables streaming free-viewpoint 3D video. By reusing renders of static regions and focusing on dynamic content, it renders at roughly 350 frames per second on Tensor Core GPUs. Why it matters: It makes real-time, multi-angle streaming feasible for live sports, virtual meetings, and robot teleoperation.
Researchers just introduced TARFlow, a scalable architecture that pairs autoregressive Transformers with exact-likelihood normalizing flows. By revamping traditional coupling layers and borrowing guidance techniques, it achieves diffusion-level quality across images and video while remaining fully invertible.
Google just unveiled Willow, a state-of-the-art quantum chip that cuts errors exponentially as physical qubits scale up. It cracked a nearly 30-year challenge in quantum error correction and finished a benchmark in under five minutes that would take supercomputers 10 septillion years. Why it matters: It proves large-scale, fault-tolerant quantum computers are actually buildable.
