Claude-style models gave different results on art interpretation
refused or stopped when asked to interpret abstract marks on paper or an art image. The same kind of task worked through on the arc.animalabs.al, with no technical setup or needed.
also worked well on a paid account for creative writing, choosing a suited to the task and producing polished text. This is a experience about how different models handled the same broad creative request.
Key points
- failed or paused on an art interpretation request.
- worked through arc.animalabs.ai without extra technical setup.
- on a paid account selected a for creative work.
- The main useful point is that refusal behavior and writing style can vary a lot by model.