embodied AI
7 published articles
Artificial Intelligence
A lens smudge just broke the AI that follows you through buildings
AdvNav is a black-box adversarial attack that uses lens-like Perlin noise to confuse vision-language navigation agents. By observing only the agent's behavior, not its internal weights, the method achieved up to 87% attack success rate. The framework reveals a broad vulnerability in current VLN systems.
2026-07-28
Embodied AI
A dozen physics engines are fighting for the right to train your robot
Robot simulation is the backbone of embodied AI training. As engines specialize and multiply, the question is no longer which one is fastest, but which abstractions will survive as shared infrastructure. NVIDIA, DeepMind, and Disney-backed Newton points toward an open, layered future where no single engine owns the stack.
2026-07-25
AI research
Alibaba's Qwen just turned world modeling into a language problem
Qwen-AgentWorld is a native language world model that simulates agent environments across seven domains. By making environment modeling the training objective from the start, it offers an alternative to video-based world models and domain-specific simulators for embodied AI.
2026-07-19
Embodied AI
Qwen's new robotics models skip the usual AI shortcut
Alibaba's Qwen team released three robotics-specific foundation models and a world model for agent environments, shifting from general-purpose vision-language models to purpose-built architectures for physical action.
2026-07-18
Embodied AI
Mistral's tiny 8B model just made lidar optional for office robots
Mistral AI's Robostral Navigate is an 8B model that uses only one RGB camera to achieve 76.6% on R2R-CE benchmarks, outperforming multi-sensor approaches by 4.5 points. Built with simulated data and a token-efficient prefix-caching method, it generalizes across robot types and adapts to unseen obstacles.
2026-07-17
Embodied AI
Xiaomi just gave every robotics lab a 38-billion-parameter data factory
With a 38-billion-parameter model that unifies four robot tasks and open-sources the entire pipeline, Xiaomi aims to break the data bottleneck in embodied AI. Early benchmarks show a 26% improvement in task completion rates when training on model-augmented data.
2026-07-15
Robotics AI
A compact robot model just beat multi-sensor systems with one camera and no depth sensors
Mistral AI launches Robostral Navigate, a compact 8B model that lets robots navigate complex indoor environments using just one camera. It beats multi-sensor approaches by 4.5 points on the R2R-CE benchmark, runs on wheeled, legged and flying robots, and was trained efficiently with prefix-caching to slash token count 22-fold.
2026-07-08