3️⃣ Adds channel-wise linear scaling inside feature blocks, boosting accuracy with minimal parameters and ZERO inference delay due to operator fusion. 4️⃣ Of course, flexible scaling on the Pareto…