Qingyu Chen

Performance gaps in multilingual medical evaluations reveal that proprietary models outperform open-source ones, but translation quality can swing results dramatically.

Retrieval strategies validated on one model can lead to significant performance gains in another, but only after rigorous experimental validation—VERDI makes this possible.