Model training
2 articles on this topic, newest first.
2026
- PublishedThe capability seesaw when AI models learn from several teachersThe P²-MOPD preprint makes balanced skill recovery the story. Here is why an average can hide a student model losing one teacher’s capability.9 min
- PublishedCan a transparent training recipe make AI progress easier to measure?Apertus 1.5 makes training stages and artifacts part of the evidence. That can help researchers distinguish a better model from a better explanation of why it improved.9 min