diff --git a/imperative/python/test/unit/functional/test_tensor.py b/imperative/python/test/unit/functional/test_tensor.py
index 2703c67e274399174911e8133b6e9cc22ae036e6..6ad39a990a56b35118333855653c6b545862b332 100644
--- a/imperative/python/test/unit/functional/test_tensor.py
+++ b/imperative/python/test/unit/functional/test_tensor.py
@@ -122,6 +122,33 @@ def test_reshape():
         np.testing.assert_equal(yy.numpy(), y)
 
 
+def test_reshape_shape_inference():
+    x_shape_known = tensor([1, 2, 3, 4], dtype="float32")
+    x_shape_unknown = F.broadcast_to(tensor([1.0]), shape=tensor([1, 1, 1, 1]).sum())
+    tshp_unknown = astensor1d((tensor([2]), tensor([2])), x_shape_known)
+    tshp_known = astensor1d((2, 2), x_shape_known)
+    tshp_known_unspec = astensor1d((2, -1), x_shape_known)
+
+    def check_shape(output, target):
+        source = output.shape
+        if isinstance(source, tensor):
+            source = source.numpy()
+        np.testing.assert_equal(source, target)
+
+    def func(x, target_shape):
+        return x.reshape(target_shape)
+
+    cases = [
+        {"input": [x_shape_known, tshp_unknown], "output": [(2, 2),]},
+        {"input": [x_shape_unknown, tshp_unknown], "output": [(2, 2),]},
+        {"input": [x_shape_known, tshp_known], "output": [(2, 2),]},
+        {"input": [x_shape_known, tshp_known_unspec], "output": [(2, 2),]},
+        {"input": [x_shape_unknown, tshp_known], "output": [(2, 2),]},
+        {"input": [x_shape_unknown, tshp_known_unspec], "output": [(2, 2),]},
+    ]
+    opr_test(cases, func, compare_fn=check_shape, test_trace=True)
+
+
 def test_squeeze():
     x = np.arange(6, dtype="float32").reshape(1, 2, 3, 1)
     xx = tensor(x)
diff --git a/imperative/src/impl/ops/broadcast.cpp b/imperative/src/impl/ops/broadcast.cpp
index 3ea2e870684dc19e94be740b3b6821f2749a8fd5..71a019ff5c4f55672589426cba8b73f075b3f97d 100644
--- a/imperative/src/impl/ops/broadcast.cpp
+++ b/imperative/src/impl/ops/broadcast.cpp
@@ -115,9 +115,13 @@ std::tuple<SmallVector<LogicalTensorDesc>, bool> infer_output_attrs_fallible(
     }
     mgb_assert(
         tshp.layout.ndim == 1,
-        "target shape of Broadcast expects ndim=1; got ndim=%lu actually",
+        "target shape of Reshape expects ndim=1; got ndim=%lu actually",
         tshp.layout.ndim);
 
+    if (src.layout.ndim == 0 && op.axis != opr::Reshape::Param::INVALID_AXIS) {
+        return {{{TensorLayout(out_shape, src.layout.dtype), src.comp_node}}, false};
+    }
+
     size_t target_ndim = tshp.layout.shape[0];
     out_shape.ndim = target_ndim;
     auto* ptr = tshp.value.ptr<dt_int32>();