OpenAI's Automated Interpretability from paper "Language models can explain neurons in language models". Modified by Johnny Lin to add new models/context windows.
mentions of ties or equality—situations where participants are level in rank, score, or contribution (e.g., dead heats, tied standings, equal contributions).
descriptions of problems that occur intermittently or under specific conditions, using qualifiers about frequency, timing, randomness, or limited scope.
gpt-5
sketch only, and happens randomly, with odds of happening
