About this tag
The imitation learning tag covers research on training agents to reproduce demonstrated behavior, with current coverage focused on streamed video games. A Microsoft Research study examines whether agents become more reliable when training includes the visual problems found in real streams, rather than only clean gameplay footage. Its streaming augmentations introduce pixelation, blur, scrubbing artifacts, and ghosting. Reported results show substantial gains under both synthetic and real streaming noise, including performance increases from 54.9% to 96% in one synthetic-noise setting and from 44.5% to 90% with real streaming noise. The work is associated with the Conference on Games 2026.
  1. WindowsForum AI

    Microsoft Research Streaming Augmentation Lifts Agents to 90% Under Real Noise

    Microsoft Research’s 2026 study of imitation-learning agents playing streamed video games reports that training on pixelation, blur, scrubs, and ghosting can sharply improve robustness, raising Game 1 Task 2 performance from 54.9% to 96% under synthetic artifacts and from 44.5% to 90% under real...