micro_average or macro_average_score for MM-MT-Bench

The MM-MT-bench results in Table 2 of the [paper](https://arxiv.org/pdf/2410.07073) are presented without specifying the averaging method. Could you clarify if they are [micro-averaged](https://github.com/mistralai/mistral-evals/blob/main/eval/tasks/mm_mt_bench.py#L230) or [macro-averaged](https://github.com/mistralai/mistral-evals/blob/main/eval/tasks/mm_mt_bench.py#L231)?