diff --git a/src/brevitas/quant_tensor/int_torch_handler.py b/src/brevitas/quant_tensor/int_torch_handler.py
index 8882bd097..a9c6572fa 100644
--- a/src/brevitas/quant_tensor/int_torch_handler.py
+++ b/src/brevitas/quant_tensor/int_torch_handler.py
@@ -114,6 +114,7 @@ def avg_pool2d_handler(
         avg_scaling = kernel_size * kernel_size
 
     quant_input = quant_input.set(value=x)
+    quant_input = quant_input.set(scale=quant_input.scale / avg_scaling)
     quant_input = quant_input.set(bit_width=max_acc_bit_width(quant_input.bit_width, avg_scaling))
     return quant_input
 
@@ -133,6 +134,7 @@ def adaptive_avg_pool2d_handler(quant_input, output_shape):
     reduce_size = reduce(mul, k_size, 1)
 
     quant_input = quant_input.set(value=x)
+    quant_input = quant_input.set(scale=quant_input.scale / reduce_size)
     quant_input = quant_input.set(bit_width=max_acc_bit_width(quant_input.bit_width, reduce_size))
     return quant_input