1 paper · 1 filter
Ashima Sood, Bryan Gardiner, Joan Condell
Jointly fine-tuning an LLM on meeting-summarisation corpora of widely varying size raises a question that prior work leaves confounded: when a domain-balanced training mixture help…