FedQUIT: On-Device Federated Unlearning via a Quasi-Competent Virtual Teacher

FedQUIT addresses the 'right to be forgotten' in Federated Learning (FL) by providing a mechanism for clients to remove their data influence from a global model. Technically, it uses a 'Quasi-Competent Virtual Teacher' to guide the distillation process, which is a clever middle ground between expensive retraining and naive fine-tuning. However, from a competitive standpoint, the project currently sits at a score of 2 because it is a very recent research release (4 days old) with zero stars and no community beyond the authors. Its primary value is as a reference implementation of a paper (arXiv:2408.07587). The moat is non-existent; the algorithm is easily reproducible by any ML engineer reading the paper. Frontier risk is medium because while OpenAI/Anthropic don't prioritize FL, Google (the pioneer of FL) has a massive vested interest in 'unlearning' for GDPR compliance in products like Gboard. If this technique proves superior to existing methods like FedEraser, it will likely be absorbed into major FL frameworks like TensorFlow Federated (TFF) or Flower, making the standalone project obsolete (High Platform Domination Risk). The displacement horizon is relatively short (1-2 years) as machine unlearning is currently a highly active research area with frequent breakthroughs.