After topping Korean MTEB without being trained on Korean ( mLateOn now tops another leaderboard that includes languages it has not been trained for and outperforms much larger models (and even rerankers)
mLateOn benches results are strong, but what's not evaluated by current benches makes it even stronger