diff --git a/.lintrunner.toml b/.lintrunner.toml
index d0c9c6aef6a..396b7fde5ac 100644
--- a/.lintrunner.toml
+++ b/.lintrunner.toml
@@ -367,7 +367,7 @@ exclude_patterns = [
     '**/third-party/**',
     'scripts/check_binary_dependencies.py',
     'profiler/test/test_profiler_e2e.py',
-    'backends/arm/test/**',
+    'backends/arm/test/ops/*.py',
 ]
 command = [
     'python',
diff --git a/.mypy.ini b/.mypy.ini
index e01392a0dfd..0ce444e8a79 100644
--- a/.mypy.ini
+++ b/.mypy.ini
@@ -24,11 +24,14 @@ files =
     test,
     util
 
-mypy_path = executorch
+mypy_path = executorch,src
 
 [mypy-executorch.backends.*]
 follow_untyped_imports = True
 
+[mypy-backends.arm.*]
+disallow_untyped_decorators = False
+
 [mypy-executorch.codegen.*]
 follow_untyped_imports = True
 
diff --git a/backends/arm/test/models/stable_diffusion/test_SD3Transformer2DModel.py b/backends/arm/test/models/stable_diffusion/test_SD3Transformer2DModel.py
index 9506fe727db..3e1f19dd39c 100644
--- a/backends/arm/test/models/stable_diffusion/test_SD3Transformer2DModel.py
+++ b/backends/arm/test/models/stable_diffusion/test_SD3Transformer2DModel.py
@@ -7,7 +7,9 @@
 from typing import Tuple
 
 import torch
-from diffusers.models.transformers import SD3Transformer2DModel
+from diffusers.models.transformers import (  # type: ignore[import-not-found]
+    SD3Transformer2DModel,
+)
 
 from executorch.backends.arm.test import common
 from executorch.backends.arm.test.models.stable_diffusion.stable_diffusion_module_test_configs import (
diff --git a/backends/arm/test/models/stable_diffusion/test_vae_AutoencoderKL.py b/backends/arm/test/models/stable_diffusion/test_vae_AutoencoderKL.py
index a3c3a018131..5d33576a817 100644
--- a/backends/arm/test/models/stable_diffusion/test_vae_AutoencoderKL.py
+++ b/backends/arm/test/models/stable_diffusion/test_vae_AutoencoderKL.py
@@ -7,8 +7,12 @@
 from typing import Tuple
 
 import torch
-from diffusers.models.autoencoders import AutoencoderKL
-from diffusers.utils.testing_utils import floats_tensor
+from diffusers.models.autoencoders import (  # type: ignore[import-not-found]
+    AutoencoderKL,
+)
+from diffusers.utils.testing_utils import (  # type: ignore[import-not-found]
+    floats_tensor,
+)
 
 from executorch.backends.arm.test import common
 from executorch.backends.arm.test.models.stable_diffusion.stable_diffusion_module_test_configs import (
diff --git a/backends/arm/test/models/test_nss.py b/backends/arm/test/models/test_nss.py
index 82187d2b479..5f7db548109 100644
--- a/backends/arm/test/models/test_nss.py
+++ b/backends/arm/test/models/test_nss.py
@@ -19,7 +19,9 @@
 
 from huggingface_hub import hf_hub_download
 
-from ng_model_gym.usecases.nss.model.model_blocks import AutoEncoderV1
+from ng_model_gym.usecases.nss.model.model_blocks import (  # type: ignore[import-not-found,import-untyped]
+    AutoEncoderV1,
+)
 
 input_t = Tuple[torch.Tensor]  # Input x
 
diff --git a/backends/arm/test/passes/test_fuse_duplicate_users_pass.py b/backends/arm/test/passes/test_fuse_duplicate_users_pass.py
index a7e80794015..ffe56e72691 100644
--- a/backends/arm/test/passes/test_fuse_duplicate_users_pass.py
+++ b/backends/arm/test/passes/test_fuse_duplicate_users_pass.py
@@ -3,7 +3,7 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
 
-from typing import Tuple
+from typing import Dict, Tuple
 
 import torch
 from executorch.backends.arm._passes import FuseDuplicateUsersPass
@@ -13,7 +13,12 @@
 input_t = Tuple[torch.Tensor]  # Input x
 
 
-class FuseaAvgPool(torch.nn.Module):
+class ModuleWithOps(torch.nn.Module):
+    ops_before_pass: Dict[str, int]
+    ops_after_pass: Dict[str, int]
+
+
+class FuseaAvgPool(ModuleWithOps):
     ops_before_pass = {
         "executorch_exir_dialects_edge__ops_aten_avg_pool2d_default": 3,
     }
@@ -27,7 +32,7 @@ def forward(self, x):
         return self.avg(x) + self.avg(x) + self.avg(x)
 
 
-class FuseAvgPoolChain(torch.nn.Module):
+class FuseAvgPoolChain(ModuleWithOps):
     ops_before_pass = {
         "executorch_exir_dialects_edge__ops_aten_avg_pool2d_default": 6,
     }
@@ -44,14 +49,14 @@ def forward(self, x):
         return first + second + third
 
 
-modules = {
+modules: Dict[str, ModuleWithOps] = {
     "fuse_avg_pool": FuseaAvgPool(),
     "fuse_avg_pool_chain": FuseAvgPoolChain(),
 }
 
 
 @common.parametrize("module", modules)
-def test_fuse_duplicate_ops_FP(module: torch.nn.Module):
+def test_fuse_duplicate_ops_FP(module: ModuleWithOps):
     pipeline = PassPipeline[input_t](
         module=module,
         test_data=(torch.ones(1, 1, 1, 1),),
diff --git a/backends/arm/test/runner_utils.py b/backends/arm/test/runner_utils.py
index 2c76af1b779..5e22a54d393 100644
--- a/backends/arm/test/runner_utils.py
+++ b/backends/arm/test/runner_utils.py
@@ -36,7 +36,7 @@
 from torch.fx.node import Node
 
 from torch.overrides import TorchFunctionMode
-from tosa.TosaGraph import TosaGraph  # type: ignore[import-untyped]
+from tosa.TosaGraph import TosaGraph  # type: ignore[import-not-found, import-untyped]
 
 logger = logging.getLogger(__name__)
 
@@ -762,7 +762,7 @@ def run_tosa_graph(
     inputs_np = [torch_tensor_to_numpy(input_tensor) for input_tensor in inputs]
 
     if isinstance(tosa_version, Tosa_1_00):
-        import tosa_reference_model as reference_model  # type: ignore[import-untyped]
+        import tosa_reference_model as reference_model  # type: ignore[import-not-found, import-untyped]
 
         debug_mode = "ALL" if logger.getEffectiveLevel() <= logging.DEBUG else None
         outputs_np, status = reference_model.run(