analiza aiqbench

The work targets a real bottleneck in robot learning: collecting dexterous manipulation data is slow and expensive, so turning a single human video into training data could cut that cost. The page content is only a title and lab attribution, so the specific method, benchmarks, and how well the synthesized data transfers to real robots can't be verified here. Readers should look for those results in the full paper before judging how much it reduces data collection effort.