Numbers/names of ports, rails during evaluation

A significant technical ambiguity has surfaced in the "AI for Industry Challenge," a competition focused on developing AI for real-world industrial tasks. A participant, driesdirckx, raised a critical issue on April 20, 2026, regarding the evaluation phase for a physical network cabling simulation. The challenge involves an AI agent correctly connecting cables to specific ports on network equipment, but the official task description notes that "the names and numbers of the ports are changed" during evaluation. This creates a major point of confusion: if multiple ports (like SC or SFP types) can appear on a single rail, how should the AI determine where to insert the cable? The core question is whether the agent must dynamically infer the correct port from the visual or environmental context, or if there is a hidden, consistent numbering order it must discover.

The confusion extends to whether rail numbers themselves remain static or also change, compounding the problem for developers trying to create a robust policy. This isn't an isolated concern, as the forum thread shows related discussions about multiple NIC/SC plugs and mount definitions, indicating systemic documentation gaps. This ambiguity strikes at the heart of the challenge's validity—if the evaluation environment's rules are not clearly defined or consistently applied, it becomes impossible to fairly judge which AI solution is truly more capable of reasoning and adapting in a physical, variable setting. The resolution of this issue will directly impact every team's ability to train and score their models, making it a pivotal moment for the competition's credibility.