[Question] Does audio-flamingo-3 support batched in-context learning?

Thank you for releasing **audio-flamingo-3** — it’s been extremely helpful in my research!

I’d like to confirm whether the model supports **batched in-context learning**.  
Concretely, I’m wondering if it is possible to provide **multiple** (audio + label) demonstration pairs in a **single** forward pass, followed by a batch of query samples.
