WillHeld

/

DiVA-llama-3-v0-8b

like 35

Hello William,

Thank you for making DiVA open-source. I've downloaded the code and am currently trying to reproduce some of the results reported in your ACL 2025 paper (https://aclanthology.org/2025.acl-long.388.pdf).

I was wondering whether it would be possible for you to share the prompts used to evaluate DiVA, particularly for the speech classification, speech translation, and spoken question answering tasks. I was also curious whether you used the same prompts across all evaluated models, or whether the prompts were tuned separately for each model reported in the paper.

Many thanks in advance, and congratulations on the work.