diff --git a/.gitignore b/.gitignore
index c29207a11..96be72089 100644
--- a/.gitignore
+++ b/.gitignore
@@ -28,3 +28,8 @@ test/*.json
 
 # venvs
 *.lock*
+dist/*
+uv.lock
+*.egg
+*.egg-info/
+*.whl
diff --git a/Makefile b/Makefile
index febe16f2f..8f7b7a034 100644
--- a/Makefile
+++ b/Makefile
@@ -71,7 +71,7 @@ clean: clean-doc clean-build clean-examples
 
 # Build a distribution in ./dist
 build:
-	$(PYTHON) setup.py sdist
+	uv build
 
 doc:
 	$(MAKE) -C ${DOCDIR} docs
@@ -82,7 +82,7 @@ doc:
 # Publish to testpypi
 # Will echo the commands to actually publish to be run to publish to actual PyPi
 # This is done to prevent accidental publishing but provide the same conveniences
-publish: clean-build build
+publish: clean build
 	$(PIP) install twine
 	$(PYTHON) -m twine upload --verbose --repository testpypi ${DIST}/*
 	@echo
@@ -90,4 +90,4 @@ publish: clean-build build
 	@echo "pip install --index-url https://test.pypi.org/simple/ dacbench"
 	@echo
 	@echo "Once you have decided it works, publish to actual pypi with"
-	@echo "python -m twine upload dist/*"
+	@echo "python -m twine upload dist/*"
\ No newline at end of file
diff --git a/dacbench/__init__.py b/dacbench/__init__.py
index 8e2224921..613596d5f 100644
--- a/dacbench/__init__.py
+++ b/dacbench/__init__.py
@@ -6,7 +6,7 @@
 from dacbench.abstract_benchmark import AbstractBenchmark
 from dacbench.abstract_env import AbstractEnv, AbstractMADACEnv
 
-__all__ = ["AbstractEnv", "AbstractMADACEnv", "AbstractBenchmark"]
+__all__ = ["AbstractBenchmark", "AbstractEnv", "AbstractMADACEnv"]
 
 from gymnasium.envs.registration import register
 
diff --git a/dacbench/abstract_env.py b/dacbench/abstract_env.py
index 123e0dc79..bb2110abe 100644
--- a/dacbench/abstract_env.py
+++ b/dacbench/abstract_env.py
@@ -190,9 +190,9 @@ def reset_(
         if scheme is None:
             scheme = options.get("scheme", self.instance_updates)
         if instance is None:
-            instance = options.get("instance", None)
+            instance = options.get("instance")
         if instance_id is None:
-            instance_id = options.get("instance_id", None)
+            instance_id = options.get("instance_id")
         self.use_next_instance(instance, instance_id, scheme=scheme)
 
     def use_next_instance(self, instance=None, instance_id=None, scheme=None):
diff --git a/dacbench/agents/__init__.py b/dacbench/agents/__init__.py
index 8211fc619..0ac4c2dd1 100644
--- a/dacbench/agents/__init__.py
+++ b/dacbench/agents/__init__.py
@@ -2,4 +2,4 @@
 from dacbench.agents.generic_agent import GenericAgent
 from dacbench.agents.simple_agents import RandomAgent, StaticAgent
 
-__all__ = ["StaticAgent", "RandomAgent", "GenericAgent", "DynamicRandomAgent"]
+__all__ = ["DynamicRandomAgent", "GenericAgent", "RandomAgent", "StaticAgent"]
diff --git a/dacbench/benchmarks/__init__.py b/dacbench/benchmarks/__init__.py
index dd55503ae..726dc1740 100644
--- a/dacbench/benchmarks/__init__.py
+++ b/dacbench/benchmarks/__init__.py
@@ -10,8 +10,8 @@
 from dacbench.benchmarks.toysgd_benchmark import ToySGDBenchmark
 
 __all__ = [
-    "LubyBenchmark",
     "FunctionApproximationBenchmark",
+    "LubyBenchmark",
     "ToySGDBenchmark",
     #    "FastDownwardBenchmark",
 ]
@@ -28,7 +28,7 @@
         "please follow the installation guide."
     )
 
-sgd_spec = importlib.util.find_spec("torch")
+sgd_spec = importlib.util.find_spec("torchvision")
 found = sgd_spec is not None
 if found:
     from dacbench.benchmarks.sgd_benchmark import SGDBenchmark
diff --git a/dacbench/benchmarks/function_approximation_benchmark.py b/dacbench/benchmarks/function_approximation_benchmark.py
index b6c383eeb..7733acab3 100644
--- a/dacbench/benchmarks/function_approximation_benchmark.py
+++ b/dacbench/benchmarks/function_approximation_benchmark.py
@@ -82,7 +82,14 @@ def __init__(self, config_path=None, config=None):
 
         for key in FUNCTION_APPROXIMATION_DEFAULTS:
             if key not in self.config:
-                self.config[key] = FUNCTION_APPROXIMATION_DEFAULTS[key]
+                if key == "observation_space_args" and "config_space" in self.config:
+                    obs_length = 1 + len(self.config["config_space"]) * 4
+                    self.config[key] = [
+                        np.array([-np.inf for _ in range(obs_length)]),
+                        np.array([np.inf for _ in range(obs_length)]),
+                    ]
+                else:
+                    self.config[key] = FUNCTION_APPROXIMATION_DEFAULTS[key]
 
     def get_environment(self):
         """Return Function Approximation env with current configuration.
@@ -197,6 +204,10 @@ def get_benchmark(self, dimension=None, seed=0):
                 "Slope (dimension 1)",
                 "Action",
             ]
+            self.config.observation_space_args = [
+                np.array([-np.inf for _ in range(4)]),
+                np.array([np.inf for _ in range(4)]),
+            ]
         if dimension == 2:
             self.config.instance_set_path = "sigmoid_2D3M_train.csv"
             self.config.test_set_path = "sigmoid_2D3M_test.csv"
@@ -220,6 +231,10 @@ def get_benchmark(self, dimension=None, seed=0):
                 "Action dim 1",
                 "Action dim 2",
             ]
+            self.config.observation_space_args = [
+                np.array([-np.inf for _ in range(7)]),
+                np.array([np.inf for _ in range(7)]),
+            ]
         if dimension == 3:
             self.config.instance_set_path = "sigmoid_3D3M_train.csv"
             self.config.test_set_path = "sigmoid_3D3M_test.csv"
@@ -250,6 +265,10 @@ def get_benchmark(self, dimension=None, seed=0):
                 "Action 2",
                 "Action 3",
             ]
+            self.config.observation_space_args = [
+                np.array([-np.inf for _ in range(10)]),
+                np.array([np.inf for _ in range(10)]),
+            ]
         if dimension == 5:
             self.config.instance_set_path = "sigmoid_5D3M_train.csv"
             self.config.test_set_path = "sigmoid_5D3M_test.csv"
@@ -294,6 +313,10 @@ def get_benchmark(self, dimension=None, seed=0):
                 "Action 4",
                 "Action 5",
             ]
+            self.config.observation_space_args = [
+                np.array([-np.inf for _ in range(16)]),
+                np.array([np.inf for _ in range(16)]),
+            ]
         self.config.seed = seed
         self.read_instance_set()
         self.read_instance_set(test=True)
diff --git a/dacbench/envs/__init__.py b/dacbench/envs/__init__.py
index f9f508d57..334f777b5 100644
--- a/dacbench/envs/__init__.py
+++ b/dacbench/envs/__init__.py
@@ -12,15 +12,15 @@
 from dacbench.envs.toysgd import ToySGDEnv, ToySGDInstance
 
 __all__ = [
-    "LubyEnv",
-    "LubyInstance",
-    "luby_gen",
     "FunctionApproximationEnv",
     "FunctionApproximationInstance",
+    "LubyEnv",
+    "LubyInstance",
+    "TheoryEnv",
     #   "FastDownwardEnv",
     "ToySGDEnv",
     "ToySGDInstance",
-    "TheoryEnv",
+    "luby_gen",
 ]
 
 modcma_spec = importlib.util.find_spec("modcma")
diff --git a/dacbench/envs/theory.py b/dacbench/envs/theory.py
index b5e872f00..a9e6e8ccc 100644
--- a/dacbench/envs/theory.py
+++ b/dacbench/envs/theory.py
@@ -224,11 +224,8 @@ def crossover(
             locs_x[locs_xprime] = False
             obj = self.get_fitness_after_crossover(xprime, locs_x, locs_xprime)
 
-            if (
-                obj not in (self.fitness, xprime.fitness)
-                or (
-                    not np.array_equal(xprime.data[locs_xprime], self.data[locs_xprime])
-                )
+            if obj not in (self.fitness, xprime.fitness) or (
+                (not np.array_equal(xprime.data[locs_xprime], self.data[locs_xprime]))
                 and (not np.array_equal(self.data[locs_x], xprime.data[locs_x]))
             ):
                 n_evals += 1
diff --git a/dacbench/wrappers/__init__.py b/dacbench/wrappers/__init__.py
index dd13698a0..0c3c78439 100644
--- a/dacbench/wrappers/__init__.py
+++ b/dacbench/wrappers/__init__.py
@@ -12,11 +12,11 @@
     "ActionFrequencyWrapper",
     "EpisodeTimeWrapper",
     "InstanceSamplingWrapper",
+    "MultiDiscreteActionWrapper",
+    "ObservationWrapper",
+    "PerformanceTrackingWrapper",
+    "PolicyProgressWrapper",
     "PolicyProgressWrapper",
     "RewardNoiseWrapper",
     "StateTrackingWrapper",
-    "PerformanceTrackingWrapper",
-    "PolicyProgressWrapper",
-    "ObservationWrapper",
-    "MultiDiscreteActionWrapper",
 ]
diff --git a/pyproject.toml b/pyproject.toml
index 80dd25cf6..9c2b8ce73 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,10 +1,10 @@
 [build-system]
-requires = ["setuptools>=70", "wheel"]
+requires = ["setuptools>=75.1.0", "wheel", "setuptools_scm>3"]
 build-backend = "setuptools.build_meta"
 
 [project]
 name = "DACBench"
-version = "0.3.0"
+version = "0.4.0"
 description = "Dynamic Algorithm Configuration Benchmark"
 readme = "README.md"
 license = { file = "LICENSE" }
@@ -27,17 +27,18 @@ classifiers = [
 ]
 requires-python = ">=3.10"
 dependencies = [
-  "gymnasium==0.29.1",
-  "imageio==2.35.1",
+  "gymnasium<=0.29.1",
+  "imageio~=2.35.1",
   "numpy==1.26.4",
   "pandas==2.2.3",
   "matplotlib==3.9.2",
   "seaborn==0.13.2",
-  "configspace==1.2.0",
+  "configspace~=1.2.1",
   "scikit-learn==1.5.2",
   "scipy==1.14.1", 
   "jupyter==1.0.0",
-  "Pyro4==4.82"]
+  "Pyro4==4.82",
+  "ioh==0.3.17"]
 
 [project.urls]
 Homepage = "https://github.com/automl/DACBench"
diff --git a/setup.py b/setup.py
index 74046109e..ca8aab328 100644
--- a/setup.py
+++ b/setup.py
@@ -19,7 +19,7 @@ def get_other_requirements():
 
 
 setup(
-    version="0.2.1",
+    version="0.3.0",
     packages=find_packages(
         exclude=[
             "tests",
diff --git a/wheels/setuptools-75.1.0-py3-none-any.whl b/wheels/setuptools-75.1.0-py3-none-any.whl
new file mode 100644
index 000000000..b7f887bd4
Binary files /dev/null and b/wheels/setuptools-75.1.0-py3-none-any.whl differ
diff --git a/wheels/wheel-0.44.0-py3-none-any.whl b/wheels/wheel-0.44.0-py3-none-any.whl
new file mode 100644
index 000000000..96431aa48
Binary files /dev/null and b/wheels/wheel-0.44.0-py3-none-any.whl differ