We compared Pi 0.5 and SmolVLA on LIBERO-Spatial under photometric stress tests. The result is hard to ignore. Pi 0.5 stays close to baseline across nearly every perturbation, even at high severity....

以机械臂操作任务为例 我们给一个VLA模型一句语言指令——“拿起碗”——然后系统性地用各种方式去改它：包括打错别字、替换同义词、加入矛盾内容、塞进无意义的句子，甚至直接换成一句“我叫Franka”。...

Previously, we asked what happens when the camera lies. Now we ask something different: What happens when the instruction lies? This post shifts entirely to the language channel. In embodied systems...

Most robustness discussions in machine learning start from a convenient fiction: The camera is a faithful observer of the world. In practice, that assumption is wrong. In embodied systems, it’s often...

Michael Lewis’s Flash Boys stayed with people because it exposed a gap between the market people thought they were trading and the one they were actually deployed into. Brad Katsuyama was trying to...

Robustness only becomes visible once the world stops behaving as expected. In deployment, inputs are imperfect, execution is delayed, hardware behaves inconsistently, and environments differ in ways...

Felix (@pewdiepie) recently set out to beat @OpenAI’s ChatGPT by fine-tuning a model that outscored GPT-4o on a benchmark (on that configuration, at that time). The interesting part wasn’t the win, it...