
New Approach Teaches Robots Dexterous Skills From Single Human Demo
Researchers demonstrate a streamlined method for training multi-fingered robots to perform complex manipulation tasks by learning from human motion.
Papers, breakthroughs, benchmarks, and the long-arc trends shaping artificial intelligence. arXiv highlights and lab announcements, distilled.

Researchers demonstrate a streamlined method for training multi-fingered robots to perform complex manipulation tasks by learning from human motion.

New theoretical framework reveals how language models learn to reason, reducing millions of parameters to interpretable low-dimensional dynamics.

A new survey reveals how large language models can develop metacognitive abilities to become more reliable and trustworthy.

Requential coding technique generates dramatically shorter representations of neural networks, offering fresh insights into model generalization and learning efficiency.

Researchers develop technique to manipulate identity representations in text-to-image systems, preserving consistency across multiple generated photos.

Researchers show pretrained multimodal models can evaluate image generation without additional training, unlocking more efficient AI development.

A developer's candid take on separating transformative language model capabilities from industry overselling.

New analysis reveals significant efficiency gaps in how leading AI coding assistants handle cache and system prompts.

A production-focused guide to fixing made-up facts in customer-facing language models

A new study reveals widespread adoption of language models among active authors, exposing a gap between researcher behavior and institutional guidelines.

Researchers warn that AI adoption threatens to homogenize inquiry methods, narrowing which questions scientists pursue.

New study reveals that standard accuracy metrics for AI image generation systems fail to catch critical numerical failures during sampling.

Researchers introduce IdeaGene-Bench to test whether large language models can trace intellectual inheritance in scientific work.

New dataset and fine-tuned model demonstrate that temporal reasoning improves how AI systems solve complex logical tasks through generated video.

A new two-stage AI model tackles the challenge of creating coherent panoramic images through geometry-aware training and a custom dataset of 1 million samples.

New training method helps autoregressive video models maintain visual coherence and motion quality over longer sequences without slowing inference.

New capability-focused benchmark addresses critical gaps in how AI systems are evaluated for practical, everyday tool use.

New method uses panoramic imaging to scale 3D neural rendering to massive outdoor environments while maintaining computational efficiency.