If you're interested in this topic, 'test time adaptation' is a special case of 'dynamic evaluation', which has a long history in NNs and is in some ways better

If you're interested in this topic, 'test time adaptation' is a special case of 'dynamic evaluation', which has a long history in NNs and is in some ways better than prompt programming or RAG. Some links: https://gwern.net/doc/ai/nn/dynamic-evaluation/index https://gwern.net/doc/ai/nn/dynamic-evaluation/index