Theory explains why training AI on one task breaks another

Original: A Local Perturbation Theory for Cross-Domain Interference and Recovery in Multi-Domain RL

Source: arxiv.org ↗