diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index 5f02444af1..3162b7ae6f 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -120,7 +120,20 @@ jobs:
             evm-version: shanghai
             evm-backend: revm
             experimental-codegen: true
-          # TODO: test experimental_codegen + -Ocodesize
+
+          - python-version: ["3.11", "311"]
+            opt-mode: codesize
+            debug: false
+            evm-version: shanghai
+            evm-backend: revm
+            experimental-codegen: true
+
+          - python-version: ["3.11", "311"]
+            opt-mode: none
+            debug: false
+            evm-version: shanghai
+            evm-backend: revm
+            experimental-codegen: true
 
           # run across other python versions. we don't really need to run all
           # modes across all python versions - one is enough
diff --git a/tests/unit/compiler/test_source_map.py b/tests/unit/compiler/test_source_map.py
index e4d5e69770..d99b546403 100644
--- a/tests/unit/compiler/test_source_map.py
+++ b/tests/unit/compiler/test_source_map.py
@@ -32,14 +32,19 @@ def foo(a: uint256) -> int128:
     """
 
 
-def test_jump_map(optimize):
+def test_jump_map(optimize, experimental_codegen):
     source_map = compile_code(TEST_CODE, output_formats=["source_map"])["source_map"]
     pos_map = source_map["pc_pos_map"]
     jump_map = source_map["pc_jump_map"]
 
     expected_jumps = 1
     if optimize == OptimizationLevel.NONE:
-        expected_jumps = 3  # some jumps get optimized out when optimizer is on
+        # some jumps which don't get optimized out when optimizer is off
+        # (slightly different behavior depending if venom pipeline is enabled):
+        if not experimental_codegen:
+            expected_jumps = 3
+        else:
+            expected_jumps = 2
 
     assert len([v for v in jump_map.values() if v == "o"]) == expected_jumps
     assert len([v for v in jump_map.values() if v == "i"]) == 2
diff --git a/vyper/ast/nodes.py b/vyper/ast/nodes.py
index ca101b61d5..b4042c75a7 100644
--- a/vyper/ast/nodes.py
+++ b/vyper/ast/nodes.py
@@ -1104,6 +1104,7 @@ def _op(self, left, right):
         # r > ln(2 ** 256) / ln(l)
         if right > math.log(decimal.Decimal(2**257)) / math.log(decimal.Decimal(left)):
             raise InvalidLiteral("Out of bounds", self)
+
         return int(left**right)
 
 
diff --git a/vyper/venom/passes/sccp/eval.py b/vyper/venom/passes/sccp/eval.py
index 8acca039c0..cd4243e998 100644
--- a/vyper/venom/passes/sccp/eval.py
+++ b/vyper/venom/passes/sccp/eval.py
@@ -1,7 +1,14 @@
 import operator
 from typing import Callable
 
-from vyper.utils import SizeLimits, evm_div, evm_mod, signed_to_unsigned, unsigned_to_signed
+from vyper.utils import (
+    SizeLimits,
+    evm_div,
+    evm_mod,
+    evm_pow,
+    signed_to_unsigned,
+    unsigned_to_signed,
+)
 from vyper.venom.basicblock import IROperand
 
 
@@ -30,9 +37,11 @@ def wrapper(ops: list[IROperand]) -> int:
 
 def _wrap_binop(operation):
     def wrapper(ops: list[IROperand]) -> int:
-        first = ops[1].value
-        second = ops[0].value
-        return (int(operation(first, second))) & SizeLimits.MAX_UINT256
+        first = _signed_to_unsigned(ops[1].value)
+        second = _signed_to_unsigned(ops[0].value)
+        ret = operation(first, second)
+        assert isinstance(ret, int)
+        return ret & SizeLimits.MAX_UINT256
 
     return wrapper
 
@@ -90,16 +99,6 @@ def _evm_not(ops: list[IROperand]) -> int:
     return SizeLimits.MAX_UINT256 ^ value
 
 
-def _evm_exp(ops: list[IROperand]) -> int:
-    base = ops[1].value
-    exponent = ops[0].value
-
-    if base == 0:
-        return 0
-
-    return pow(base, exponent, SizeLimits.CEILING_UINT256)
-
-
 ARITHMETIC_OPS: dict[str, Callable[[list[IROperand]], int]] = {
     "add": _wrap_binop(operator.add),
     "sub": _wrap_binop(operator.sub),
@@ -108,7 +107,7 @@ def _evm_exp(ops: list[IROperand]) -> int:
     "sdiv": _wrap_signed_binop(evm_div),
     "mod": _wrap_binop(evm_mod),
     "smod": _wrap_signed_binop(evm_mod),
-    "exp": _evm_exp,
+    "exp": _wrap_binop(evm_pow),
     "eq": _wrap_binop(operator.eq),
     "ne": _wrap_binop(operator.ne),
     "lt": _wrap_binop(operator.lt),