Benchmark
MMEB-V2
A multimodal embedding evaluation benchmark comprising 78 tasks across 12 task groups, used to assess the performance of multimodal embedding models on various retrieval and understanding tasks.
Timeline (last 90 days)
Oct 26
On the 78 tasks of the MMEB-V2 benchmark, PILOT outperforms the main baselines at both the 2B and 7B parameter scales.
Expired50%
Oct 26
PILOT improves on all 12 task groups of MMEB-V2, raising the overall score by 3.2 and 3.3 points respectively (2B and 7B scales).
Expired50%