[BUG]: AI annotation placed on wrong chart with small multiples ensemble variable charts

[jryu01]

The annotation should be placed in the second (blue) chart.

[jryu01]

It's not a bug. If more context is provided in the instruction, e.g. model name, it works as expected.

It looks like LLM isn't just smart enough to figure out the correct chart if there are some ambiguity.
For example, in above case, actual variable name for the first chart is infected_state while the label is 'Ensemble after calibration' and llm tends to choose that over the second chart even though correct label to variable name translation is provided.

If you ask it to draw on SIDARTH, it works fine.
Closing this since I think it isn't worth fine tuning the prompt for this edge case.