diff --git a/README.md b/README.md
index 249f142..b38a354 100644
--- a/README.md
+++ b/README.md
@@ -30,7 +30,6 @@
 [![Conda](https://img.shields.io/conda/v/conda-forge/safe-mol?label=conda&color=success)](https://anaconda.org/conda-forge/safe-mol)
 [![PyPI - Downloads](https://img.shields.io/pypi/dm/safe-mol)](https://pypi.org/project/safe-mol/)
 [![Conda](https://img.shields.io/conda/dn/conda-forge/safe-mol)](https://anaconda.org/conda-forge/safe-mol)
-[![PyPI - Python Version](https://img.shields.io/pypi/pyversions/safe-mol)](https://pypi.org/project/safe-mol/)
 [![Code license](https://img.shields.io/badge/Code%20License-Apache_2.0-green.svg)](https://github.com/datamol-io/safe/blob/main/LICENSE)
 [![Data License](https://img.shields.io/badge/Data%20License-CC%20BY%204.0-red.svg)](https://github.com/datamol-io/safe/blob/main/DATA_LICENSE)
 [![GitHub Repo stars](https://img.shields.io/github/stars/datamol-io/safe)](https://github.com/datamol-io/safe/stargazers)
@@ -60,6 +59,13 @@ The construction of a SAFE strings requires defining a molecular fragmentation a
     <img src="docs/assets/safe-construction.svg" width="100%">
 </div>
 
+## News 
+
+#### 2024/01/15
+1. We have updated the model with the version used for the paper. The revision number is ``
+2. @IanAWatson has a C++ implementation of SAFE in [LillyMol](https://github.com/IanAWatson/LillyMol/tree/bazel_version_float) that is quite fast and use a custom fragmentation algorithm. Follow the installation instruction on the repo and checkout the docs of the CLI here: [docs/Molecule_Tools/SAFE.md](https://github.com/IanAWatson/LillyMol/blob/bazel_version_float/docs/Molecule_Tools/SAFE.md)
+
+
 ### Installation
 
 You can install `safe` using pip:
diff --git a/safe/__init__.py b/safe/__init__.py
index b58b0cf..4bbdbc3 100644
--- a/safe/__init__.py
+++ b/safe/__init__.py
@@ -1,12 +1,6 @@
-from .converter import encode
-from .converter import decode
-from .converter import SAFEConverter
-from .viz import to_image
-from .tokenizer import SAFETokenizer
-from .tokenizer import split
+from . import trainer, utils
+from ._exception import SAFEDecodeError, SAFEEncodeError, SAFEFragmentationError
+from .converter import SAFEConverter, decode, encode
 from .sample import SAFEDesign
-from ._exception import SAFEDecodeError
-from ._exception import SAFEEncodeError
-from ._exception import SAFEFragmentationError
-from . import trainer
-from . import utils
+from .tokenizer import SAFETokenizer, split
+from .viz import to_image
diff --git a/safe/converter.py b/safe/converter.py
index b2b2f31..7611168 100644
--- a/safe/converter.py
+++ b/safe/converter.py
@@ -1,21 +1,15 @@
-from typing import Union
-from typing import Optional
-from typing import List
-from typing import Callable
-
-import re
-import datamol as dm
 import itertools
-import numpy as np
-
-from contextlib import suppress
+import re
 from collections import Counter
+from contextlib import suppress
+from typing import Callable, List, Optional, Union
 
+import datamol as dm
+import numpy as np
 from rdkit import Chem
 from rdkit.Chem import BRICS
-from ._exception import SAFEDecodeError
-from ._exception import SAFEEncodeError
-from ._exception import SAFEFragmentationError
+
+from ._exception import SAFEDecodeError, SAFEEncodeError, SAFEFragmentationError
 from .utils import standardize_attach
 
 
@@ -110,8 +104,8 @@ def _find_branch_number(cls, inp: str):
         Args:
             inp: input smiles
         """
-
-        matching_groups = re.findall(r"((?<=%)\d{2})|((?<!%)\d+)", inp)
+        inp = re.sub("[\[].*?[\]]", "", inp)  # noqa
+        matching_groups = re.findall(r"((?<=%)\d{2})|((?<!%)\d+)(?![^\[]*\])", inp)
         # first match is for multiple connection as multiple digits
         # second match is for single connections requiring 2 digits
         # SMILES does not support triple digits
@@ -262,6 +256,7 @@ def encoder(
         # TODO(maclandrol): RDKit supports some extended form of ring closure, up to 5 digits
         # https://www.rdkit.org/docs/RDKit_Book.html#ring-closures and I should try to include them
         branch_numbers = self._find_branch_number(inp)
+
         mol = dm.to_mol(inp, remove_hs=False)
 
         bond_map_id = 1
@@ -327,7 +322,9 @@ def encoder(
             )
 
         scaffold_str = ".".join(frags_str)
-        attach_pos = set(re.findall(r"(\[\d+\*\]|\[[^:]*:\d+\])", scaffold_str))
+        # don't capture atom mapping in the scaffold
+        attach_pos = set(re.findall(r"(\[\d+\*\]|!\[[^:]*:\d+\])", scaffold_str))
+
         if canonical:
             attach_pos = sorted(attach_pos)
         starting_num = 1 if len(branch_numbers) == 0 else max(branch_numbers) + 1
diff --git a/safe/sample.py b/safe/sample.py
index f3fb1a0..064cebf 100644
--- a/safe/sample.py
+++ b/safe/sample.py
@@ -1,26 +1,22 @@
-from typing import Union
-from typing import List
-from typing import Optional
-
-
-from contextlib import suppress
-from collections.abc import Mapping
-from collections import Counter
-from transformers.generation import PhrasalConstraint
-from transformers.generation import DisjunctiveConstraint
-from transformers import GenerationConfig
-from safe.trainer.model import SAFEDoubleHeadsModel
-from safe.tokenizer import SAFETokenizer
-from loguru import logger
-from tqdm.auto import tqdm
-
 import itertools
 import os
-import re
-import torch
 import random
+import re
+from collections import Counter
+from collections.abc import Mapping
+from contextlib import suppress
+from typing import List, Optional, Union
+
 import datamol as dm
+import torch
+from loguru import logger
+from tqdm.auto import tqdm
+from transformers import GenerationConfig
+from transformers.generation import DisjunctiveConstraint, PhrasalConstraint
+
 import safe as sf
+from safe.tokenizer import SAFETokenizer
+from safe.trainer.model import SAFEDoubleHeadsModel
 
 
 class SAFEDesign:
diff --git a/safe/tokenizer.py b/safe/tokenizer.py
index 9932c3f..5f58009 100644
--- a/safe/tokenizer.py
+++ b/safe/tokenizer.py
@@ -1,41 +1,35 @@
-from typing import Optional
-from typing import List
-from typing import Any
-from typing import Iterator
-from typing import Union
-from typing import Dict
-
-import re
-import os
 import contextlib
-import fsspec
 import copy
-import torch
-import numpy as np
 import json
+import os
+import re
 import warnings
-import packaging.version
+from typing import Any, Dict, Iterator, List, Optional, Union
 
+import fsspec
+import numpy as np
+import packaging.version
+import torch
 from loguru import logger
-from tokenizers import decoders
-from tokenizers import Tokenizer
+from tokenizers import Tokenizer, decoders
 from tokenizers.models import BPE, WordLevel
-from tokenizers.trainers import BpeTrainer, WordLevelTrainer
-from tokenizers.pre_tokenizers import Whitespace, PreTokenizer
+from tokenizers.pre_tokenizers import PreTokenizer, Whitespace
 from tokenizers.processors import TemplateProcessing
+from tokenizers.trainers import BpeTrainer, WordLevelTrainer
 from transformers import PreTrainedTokenizerFast
 from transformers import __version__ as transformers_version
-from transformers.utils import PushToHubMixin
-from transformers.utils import is_offline_mode
-from transformers.utils import is_remote_url
-from transformers.utils import cached_file
-from transformers.utils import download_url
-from transformers.utils import extract_commit_hash
-from transformers.utils import working_or_temp_dir
+from transformers.utils import (
+    PushToHubMixin,
+    cached_file,
+    download_url,
+    extract_commit_hash,
+    is_offline_mode,
+    is_remote_url,
+    working_or_temp_dir,
+)
 
 from .utils import attr_as
 
-
 SPECIAL_TOKENS = ["[UNK]", "[CLS]", "[SEP]", "[PAD]", "[MASK]"]
 UNK_TOKEN = "[UNK]"
 PADDING_TOKEN = "[PAD]"
diff --git a/safe/trainer/cli.py b/safe/trainer/cli.py
index 53abd19..2b2fa69 100644
--- a/safe/trainer/cli.py
+++ b/safe/trainer/cli.py
@@ -1,30 +1,26 @@
-from typing import Optional
-from typing import Literal
-
 import math
 import os
 import sys
 import uuid
-import safe
+from dataclasses import dataclass, field
+from typing import Literal, Optional
+
+import datasets
+import evaluate
 import torch
 import transformers
-import evaluate
-import datasets
-from dataclasses import dataclass, field
 from loguru import logger
-from transformers import AutoConfig
-from transformers import AutoTokenizer
-from transformers import set_seed
-from transformers.utils.logging import log_levels as LOG_LEVELS
+from transformers import AutoConfig, AutoTokenizer, TrainingArguments, set_seed
 from transformers.trainer_utils import get_last_checkpoint
-from transformers import TrainingArguments
-from safe.trainer.model import SAFEDoubleHeadsModel
+from transformers.utils.logging import log_levels as LOG_LEVELS
+
+import safe
 from safe.tokenizer import SAFETokenizer
-from safe.trainer.data_utils import get_dataset
 from safe.trainer.collator import SAFECollator
+from safe.trainer.data_utils import get_dataset
+from safe.trainer.model import SAFEDoubleHeadsModel
 from safe.trainer.trainer_utils import SAFETrainer
 
-
 CURRENT_DIR = os.path.join(safe.__path__[0], "trainer")
 
 
diff --git a/safe/trainer/collator.py b/safe/trainer/collator.py
index 05a67b1..55b73d4 100644
--- a/safe/trainer/collator.py
+++ b/safe/trainer/collator.py
@@ -1,18 +1,13 @@
-from typing import Union
-from typing import Optional
-from typing import List
-from typing import Dict
-from typing import Any
-
 import copy
 import functools
-import torch
 from collections.abc import Mapping
+from typing import Any, Dict, List, Optional, Union
 
+import torch
+from tokenizers import Tokenizer
 from transformers.data.data_collator import _torch_collate_batch
-from safe.tokenizer import SAFETokenizer
 
-from tokenizers import Tokenizer
+from safe.tokenizer import SAFETokenizer
 
 
 class SAFECollator:
diff --git a/safe/trainer/data_utils.py b/safe/trainer/data_utils.py
index eb1e706..fa57e2a 100644
--- a/safe/trainer/data_utils.py
+++ b/safe/trainer/data_utils.py
@@ -1,16 +1,11 @@
-from typing import Optional
-from typing import Callable
-from typing import Any
-from typing import Union
-from typing import Dict
-
+import itertools
 from collections.abc import Mapping
-from tqdm.auto import tqdm
 from functools import partial
+from typing import Any, Callable, Dict, Optional, Union
 
-import itertools
-import upath
 import datasets
+import upath
+from tqdm.auto import tqdm
 
 from safe.tokenizer import SAFETokenizer
 
diff --git a/safe/trainer/model.py b/safe/trainer/model.py
index 5db6509..265bae2 100644
--- a/safe/trainer/model.py
+++ b/safe/trainer/model.py
@@ -1,21 +1,17 @@
-from typing import Optional
-from typing import Union
-from typing import Tuple
-from typing import Callable
-from typing import Any
+from typing import Any, Callable, Optional, Tuple, Union
 
 import torch
 from torch import nn
-from torch.nn import CrossEntropyLoss
-from torch.nn import MSELoss
-from transformers import GPT2DoubleHeadsModel
+from torch.nn import CrossEntropyLoss, MSELoss
+from transformers import GPT2DoubleHeadsModel, PretrainedConfig
 from transformers.activations import get_activation
-from transformers.models.gpt2.modeling_gpt2 import GPT2DoubleHeadsModelOutput
-from transformers.models.gpt2.modeling_gpt2 import add_start_docstrings_to_model_forward
-from transformers.models.gpt2.modeling_gpt2 import replace_return_docstrings
-from transformers.models.gpt2.modeling_gpt2 import GPT2_INPUTS_DOCSTRING
-from transformers.models.gpt2.modeling_gpt2 import _CONFIG_FOR_DOC
-from transformers import PretrainedConfig
+from transformers.models.gpt2.modeling_gpt2 import (
+    _CONFIG_FOR_DOC,
+    GPT2_INPUTS_DOCSTRING,
+    GPT2DoubleHeadsModelOutput,
+    add_start_docstrings_to_model_forward,
+    replace_return_docstrings,
+)
 
 
 class PropertyHead(torch.nn.Module):
diff --git a/safe/trainer/trainer_utils.py b/safe/trainer/trainer_utils.py
index f883c5e..0c5afd7 100644
--- a/safe/trainer/trainer_utils.py
+++ b/safe/trainer/trainer_utils.py
@@ -1,8 +1,6 @@
 from transformers import Trainer
 from transformers.modeling_utils import unwrap_model
-from transformers.models.auto.modeling_auto import (
-    MODEL_FOR_CAUSAL_LM_MAPPING_NAMES,
-)
+from transformers.models.auto.modeling_auto import MODEL_FOR_CAUSAL_LM_MAPPING_NAMES
 
 
 class SAFETrainer(Trainer):
diff --git a/safe/utils.py b/safe/utils.py
index 8445a02..05a4faf 100644
--- a/safe/utils.py
+++ b/safe/utils.py
@@ -1,30 +1,22 @@
-from typing import Optional
-from typing import Any
-from typing import Union
-from typing import List
-from typing import Tuple
-
+import itertools
+import random
+import re
 from collections import deque
 from contextlib import contextmanager, suppress
 from functools import partial
-from itertools import combinations
-from itertools import compress
+from itertools import combinations, compress
+from typing import Any, List, Optional, Tuple, Union
+
+import datamol as dm
+import networkx as nx
+import numpy as np
 from loguru import logger
 from networkx.utils import py_random_state
-
 from rdkit import Chem
-from rdkit.Chem import EditableMol, Atom
-from rdkit.Chem.rdmolops import ReplaceCore
-from rdkit.Chem.rdmolops import AdjustQueryParameters
-from rdkit.Chem.rdmolops import AdjustQueryProperties
+from rdkit.Chem import Atom, EditableMol
 from rdkit.Chem.rdChemReactions import ReactionFromSmarts
+from rdkit.Chem.rdmolops import AdjustQueryParameters, AdjustQueryProperties, ReplaceCore
 
-import itertools
-import random
-import re
-import numpy as np
-import networkx as nx
-import datamol as dm
 import safe as sf
 
 __implicit_carbon_query = dm.from_smarts("[#6;h]")
diff --git a/safe/viz.py b/safe/viz.py
index 369fb9f..8021d73 100644
--- a/safe/viz.py
+++ b/safe/viz.py
@@ -1,11 +1,9 @@
-from typing import Optional
-from typing import Any
-from typing import Tuple
-from typing import Union
-
 import itertools
-import matplotlib.pyplot as plt
+from typing import Any, Optional, Tuple, Union
+
 import datamol as dm
+import matplotlib.pyplot as plt
+
 import safe as sf
 
 
diff --git a/tests/test_hgf_load.py b/tests/test_hgf_load.py
index 146f373..3cf5ed2 100644
--- a/tests/test_hgf_load.py
+++ b/tests/test_hgf_load.py
@@ -1,6 +1,6 @@
+from safe.sample import SAFEDesign
 from safe.tokenizer import SAFETokenizer
 from safe.trainer.model import SAFEDoubleHeadsModel
-from safe.sample import SAFEDesign
 
 
 def test_load_default_safe_model():
diff --git a/tests/test_notebooks.py b/tests/test_notebooks.py
index 23a0ab5..802135d 100644
--- a/tests/test_notebooks.py
+++ b/tests/test_notebooks.py
@@ -1,10 +1,9 @@
-import pytest
 import pathlib
 
 import nbformat
+import pytest
 from nbconvert.preprocessors.execute import ExecutePreprocessor
 
-
 ROOT_DIR = pathlib.Path(__file__).parent.resolve()
 
 TUTORIALS_DIR = ROOT_DIR.parent / "docs" / "tutorials"
diff --git a/tests/test_safe.py b/tests/test_safe.py
index 2b3057f..fdfc288 100644
--- a/tests/test_safe.py
+++ b/tests/test_safe.py
@@ -1,7 +1,8 @@
-import pytest
 import datamol as dm
-import safe
 import numpy as np
+import pytest
+
+import safe
 
 
 def test_safe_encoding():
@@ -74,3 +75,38 @@ def test_rdkit_smiles_parser_issues():
     assert safe.decode(failing_encoded) is None
     assert working_decoded is not None
     assert dm.same_mol(working_no_stero, input_mol)
+
+
+@pytest.mark.parametrize(
+    "input_sm",
+    [
+        "O=C(CN1CC[NH2+]CC1)N1CCCCC1",
+        "[NH3+]Cc1ccccc1",
+        "c1cc2c(cc1[C@@H]1CCC[NH2+]1)OCCO2",
+        "[13C]1CCCCC1C[238U]C[NH3+]",
+        "COC[CH2:1][CH2:2]O[CH:2]C[OH:3]",
+    ],
+)
+def test_bracket_smiles_issues(input_sm):
+    slicer = "brics"
+    safe_obj = safe.SAFEConverter(slicer=slicer, require_hs=False)
+    fragments = []
+    with dm.without_rdkit_log():
+        safe_str = safe_obj.encoder(
+            input_sm,
+            canonical=True,
+        )
+        for fragment in safe_str.split("."):
+            f = safe_obj.decoder(
+                fragment,
+                as_mol=False,
+                canonical=True,
+                fix=True,
+                remove_dummies=True,
+                remove_added_hs=True,
+            )
+            fragments.append(f)
+    input_mol = dm.to_mol(input_sm)
+    assert safe.decode(safe_str) is not None
+    assert dm.same_mol(dm.to_mol(safe_str), input_mol)
+    assert None not in fragments