AI RESEARCH
Cross-Domain Demo-to-Code via Neurosymbolic Counterfactual Reasoning
arXiv CS.AI
•
ArXi:2603.18495v1 Announce Type: new Recent advances in Vision-Language Models (VLMs) have enabled video-instructed robotic programming, allowing agents to interpret video nstrations and generate executable control code. We formulate video-instructed robotic programming as a cross-domain adaptation problem, where perceptual and physical differences between nstration and deployment induce procedural mismatches. However, current VLMs lack the procedural understanding needed to reformulate causal dependencies and achieve task-compatible behavior under such domain shifts. We