Papers1061

RoboMME is a new, big test playground that checks whether robot brains can remember important things over time, not just what they see right now.

Helios is a 14-billion-parameter video model that can make minute-long videos in real time at about 19.5 frames per second on a single NVIDIA H100 GPU.

The paper shows that when a model compares two of its own answers head-to-head, it picks the right one more often than when it judges each answer alone.

CubeComposer is a new AI method that turns a normal forward-facing video into a full 360° VR video at true 4K quality without using super-resolution upscaling.

RIVER Bench is a new test that checks how well AI can watch a video stream and talk with you in real time.

Phi-4-reasoning-vision-15B is a small, open-weight AI that understands pictures and text together and is especially good at math, science, and using computer screens.

SWE-CI is a new benchmark that tests how well AI coding agents can keep a codebase healthy over many changes, not just fix one bug.

This paper shows that teaching AI to first draw a simple map of a text (nodes and links) before answering questions makes it smarter and more reliable.

Scientists want AI to propose brand‑new hypotheses directly from a research background, but training a model to do this end‑to‑end is mathematically intractable because the search space explodes combinatorially.

InfinityStory is a new system that can make very long videos (even hours) where the world stays the same and characters transition smoothly between shots.

Utonia is a single brain (encoder) that learns from many kinds of 3D point clouds, like indoor rooms, outdoor streets, tiny toys, and even city maps.

MIBURI is a system that makes a talking digital character move its body and face expressively in real time while it speaks.