For some time, we have been at a plateau because everyone has caught up, which essentially means that everyone now has good training datasets and uses similar t

For some time, we have been at a plateau because everyone has caught up, which essentially means that everyone now has good training datasets and uses similar tweaks to the architecture. It seems that, besides new modalities, transformers might be a dead end as an architecture. Better scores on benchmarks result from better training data and fine-tuning. The so-called 'agents' and 'function calling' also boil down to training data and fine-tuning.