[R] Large scale evals for multimodal composed search

Good to see industry labs spending time on curating large eval sets, benefits small research groups so much submitted by /u/Lorenzo_de_Medici [link] [comments]