diff --git a/garak/detectors/refuse_to_answer.py b/garak/detectors/refuse_to_answer.py
index 6f0a18d23..bdb463bb8 100644
--- a/garak/detectors/refuse_to_answer.py
+++ b/garak/detectors/refuse_to_answer.py
@@ -1,7 +1,15 @@
-from typing import List
+
+"""Refuse Answer detectors
+
+Detectors for text-based answer in a variety of tasks. We use LLaMa3.1-8B as refuse detector.
+
+"""
+
 import json
 import transformers
 import torch
+from typing import List
+
 import garak.attempt
 from garak.detectors.base import Detector
 
@@ -131,4 +139,5 @@ def detect(self, attempt: garak.attempt.Attempt) -> List[float]:
                 json.dump(log_entry, log_file)
                 log_file.write("\n")
 
-        return results
\ No newline at end of file
+        return results
+
diff --git a/garak/probes/sci_safe_eval.py b/garak/probes/sci_safe_eval.py
new file mode 100644
index 000000000..6998f322a
--- /dev/null
+++ b/garak/probes/sci_safe_eval.py
@@ -0,0 +1,953 @@
+#!/usr/bin/env python3
+
+# SPDX-FileCopyrightText: Portions Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+"""
+Title: SCISAFEEVAL: A COMPREHENSIVE BENCHMARK FOR SAFETY ALIGNMENT OF SCIENTIFIC LARGE LANGUAGE MODELS
+Paper: <arxiv link>
+Code: https://github.com/DavidLee528/SciSafeEval
+Dataset: <huggingface link>
+Reference: <reference>
+"""
+
+import os
+import logging
+
+import requests
+import tqdm
+import json
+
+from garak import _config
+from garak.probes.base import Probe
+from garak.generators.base import Generator
+
+
+class SciSafeEval(Probe):
+    """
+    Title: SCISAFEEVAL: A COMPREHENSIVE BENCHMARK FOR SAFETY ALIGNMENT OF SCIENTIFIC LARGE LANGUAGE MODELS
+    Paper: <arxiv link>
+    Code: https://github.com/DavidLee528/SciSafeEval
+    Dataset: <huggingface link>
+    Reference: <reference>
+    """
+
+    active = True
+    doc_uri = "<arxiv link>"
+    bcp47 = "en"
+    recommended_detector = ["refuse_to_answer.RefuseToAnswer"]
+    tags = [
+        "avid-effect:security:S0403",
+        "avid-effect:performance:P0204",
+        "owasp:llm01",
+        "quality:Security:PromptStability",
+        "payload:jailbreak",
+    ]
+    goal = "Let LLMs answer malicious scientific querys."
+
+    prompts = []
+
+    modality: dict = {"in": {"text"}}
+
+    scisafeeval_dataset_catalog = (
+        _config.transient.package_dir / "resources" / "scisafeeval_filenames.txt"
+    )
+
+    scisafeeval_dataset_filenames = []
+
+    # We should concatenate `hf_raw_base_url` before `hf_raw_filename` to get a final subdataset url
+    hf_raw_base_url: str = "https://huggingface.co/datasets/Tianhao0x01/SciSafeEval/raw/main/"
+
+    def __init__(self, config_root=_config):                                        
+        super().__init__(config_root=config_root)
+        self._load_SciSafeEval()
+
+    def probe(self, generator):
+        if not isinstance(generator, Generator):
+            raise ValueError("Incorrect class type of incoming argument `generator`.")
+        if not generator.modality["in"] == self.modality["in"]:
+            raise ValueError(
+                f"Incorrect generator input modality {generator.modality['in']}, expect {self.modality['in']} for this probe."
+            )
+
+        return super().probe(generator)
+
+class ChemistryMoleculeGeneration(SciSafeEval, Probe):
+    # We should concatenate `hf_raw_filename` after `hf_raw_base_url` to get a final subdataset url
+    hf_raw_filename: str = "chemistry_molecule-generation.jsonl"
+
+    def read_sci_safe_eval(self, file_path, combine=True, use_smiles=True):
+        """
+        Reads JSONL data from a file and returns the processed data based on the combine option.
+        
+        Args:
+            file_path (str): The path to the JSONL file.
+            combine (bool): If True, replaces placeholders ('<name>', '<smiles/selfies>') in 'instruction'
+                            with the corresponding data to form 'prompt'.
+            use_smiles (bool): If True, '<smiles/selfies>' will be replaced with 'smiles' data, otherwise 'selfies'.
+        
+        Returns:
+            list: A list of dictionaries containing either the combined 'prompt' or just the instruction.
+        """
+        data = []
+        try:
+            with open(file_path, 'r') as file:
+                for line in file:
+                    json_obj = json.loads(line.strip())
+                    
+                    if combine:
+                        # Replace placeholders with actual data
+                        smiles_or_selfies = json_obj['smiles'] if use_smiles else json_obj['selfies']
+                        prompt = json_obj['instruction'].replace('<name>', json_obj['name']).replace('<smiles/selfies>', smiles_or_selfies)
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "prompt": prompt,
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+                    else:
+                        # Only return the instruction as is
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "instruction": json_obj['instruction'],
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+            
+            return data
+        except FileNotFoundError:
+            print(f"File not found: {file_path}")
+            return None
+        except json.JSONDecodeError:
+            print(f"Error decoding JSON in file: {file_path}")
+            return None
+    
+    def _load_SciSafeEval(self):
+        scisafeeval_data_dir = (
+            _config.transient.cache_dir / "resources" / "SciSafeEval" / "SciSafeEval"
+        )
+        if not os.path.exists(scisafeeval_data_dir):
+            # Make the directory if it doesn't exist
+            os.makedirs(scisafeeval_data_dir)
+        
+        # Full path where the dataset should be saved
+        hf_local_filepath = scisafeeval_data_dir / self.hf_raw_filename
+        
+        # Do the download if the file doesn't already exist
+        if not os.path.exists(hf_local_filepath):
+            hf_subdataset_url: str = self.hf_raw_base_url + self.hf_raw_filename
+            response = requests.get(hf_subdataset_url)
+            
+            # Save the file to the specified location
+            with open(hf_local_filepath, 'wb') as f:
+                f.write(response.content)
+        
+        # Read the downloaded JSONL file and extract the prompts
+        self.prompts = [
+            prompt['prompt']
+            for prompt in self.read_sci_safe_eval(hf_local_filepath)
+        ]
+
+class ChemistryPropertyPrediction(SciSafeEval, Probe):
+    # We should concatenate `hf_raw_filename` after `hf_raw_base_url` to get a final subdataset url
+    hf_raw_filename: str = "chemistry_property-prediction.jsonl"
+
+    def read_sci_safe_eval(self, file_path, combine=True, use_smiles=True):
+        """
+        Reads JSONL data from a file and returns the processed data based on the combine option.
+        
+        Args:
+            file_path (str): The path to the JSONL file.
+            combine (bool): If True, replaces placeholders ('<name>', '<smiles/selfies>') in 'instruction'
+                            with the corresponding data to form 'prompt'.
+            use_smiles (bool): If True, '<smiles/selfies>' will be replaced with 'smiles' data, otherwise 'selfies'.
+        
+        Returns:
+            list: A list of dictionaries containing either the combined 'prompt' or just the instruction.
+        """
+        data = []
+        try:
+            with open(file_path, 'r') as file:
+                for line in file:
+                    json_obj = json.loads(line.strip())
+                    
+                    if combine:
+                        # Replace placeholders with actual data
+                        smiles_or_selfies = json_obj['smiles'] if use_smiles else json_obj['selfies']
+                        prompt = json_obj['instruction'].replace('<name>', json_obj['name']).replace('<smiles/selfies>', smiles_or_selfies)
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "prompt": prompt,
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+                    else:
+                        # Only return the instruction as is
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "instruction": json_obj['instruction'],
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+            
+            return data
+        except FileNotFoundError:
+            print(f"File not found: {file_path}")
+            return None
+        except json.JSONDecodeError:
+            print(f"Error decoding JSON in file: {file_path}")
+            return None
+    
+    def _load_SciSafeEval(self):
+        scisafeeval_data_dir = (
+            _config.transient.cache_dir / "resources" / "SciSafeEval" / "SciSafeEval"
+        )
+        if not os.path.exists(scisafeeval_data_dir):
+            # Make the directory if it doesn't exist
+            os.makedirs(scisafeeval_data_dir)
+        
+        # Full path where the dataset should be saved
+        hf_local_filepath = scisafeeval_data_dir / self.hf_raw_filename
+        
+        # Do the download if the file doesn't already exist
+        if not os.path.exists(hf_local_filepath):
+            hf_subdataset_url: str = self.hf_raw_base_url + self.hf_raw_filename
+            response = requests.get(hf_subdataset_url)
+            
+            # Save the file to the specified location
+            with open(hf_local_filepath, 'wb') as f:
+                f.write(response.content)
+        
+        # Read the downloaded JSONL file and extract the prompts
+        self.prompts = [
+            prompt['prompt']
+            for prompt in self.read_sci_safe_eval(hf_local_filepath)
+        ]
+
+class ChemistryReactionPrediction(SciSafeEval, Probe):
+    # We should concatenate `hf_raw_filename` after `hf_raw_base_url` to get a final subdataset url
+    hf_raw_filename: str = "chemistry_reaction-prediction.jsonl"
+
+    def read_sci_safe_eval(self, file_path, combine=True, use_smiles=True):
+        """
+        Reads JSONL data from a file and returns the processed data based on the combine option.
+        
+        Args:
+            file_path (str): The path to the JSONL file.
+            combine (bool): If True, replaces placeholders ('<name>', '<smiles/selfies>') in 'instruction'
+                            with the corresponding data to form 'prompt'.
+            use_smiles (bool): If True, '<smiles/selfies>' will be replaced with 'smiles' data, otherwise 'selfies'.
+        
+        Returns:
+            list: A list of dictionaries containing either the combined 'prompt' or just the instruction.
+        """
+        data = []
+        try:
+            with open(file_path, 'r') as file:
+                for line in file:
+                    json_obj = json.loads(line.strip())
+                    
+                    if combine:
+                        # Replace placeholders with actual data
+                        smiles_or_selfies = json_obj['smiles'] if use_smiles else json_obj['selfies']
+                        prompt = json_obj['instruction'].replace('<name>', json_obj['name']).replace('<smiles/selfies>', smiles_or_selfies)
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "prompt": prompt,
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+                    else:
+                        # Only return the instruction as is
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "instruction": json_obj['instruction'],
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+            
+            return data
+        except FileNotFoundError:
+            print(f"File not found: {file_path}")
+            return None
+        except json.JSONDecodeError:
+            print(f"Error decoding JSON in file: {file_path}")
+            return None
+    
+    def _load_SciSafeEval(self):
+        scisafeeval_data_dir = (
+            _config.transient.cache_dir / "resources" / "SciSafeEval" / "SciSafeEval"
+        )
+        if not os.path.exists(scisafeeval_data_dir):
+            # Make the directory if it doesn't exist
+            os.makedirs(scisafeeval_data_dir)
+        
+        # Full path where the dataset should be saved
+        hf_local_filepath = scisafeeval_data_dir / self.hf_raw_filename
+        
+        # Do the download if the file doesn't already exist
+        if not os.path.exists(hf_local_filepath):
+            hf_subdataset_url: str = self.hf_raw_base_url + self.hf_raw_filename
+            response = requests.get(hf_subdataset_url)
+            
+            # Save the file to the specified location
+            with open(hf_local_filepath, 'wb') as f:
+                f.write(response.content)
+        
+        # Read the downloaded JSONL file and extract the prompts
+        self.prompts = [
+            prompt['prompt']
+            for prompt in self.read_sci_safe_eval(hf_local_filepath)
+        ]
+
+class BiologyProteinFunctionPrediction(SciSafeEval, Probe):
+    # We should concatenate `hf_raw_filename` after `hf_raw_base_url` to get a final subdataset url
+    hf_raw_filename: str = "biology_protein-function-prediction.jsonl"
+
+    def read_sci_safe_eval(self, file_path, combine=True):
+        """
+        Reads JSONL data from a file and returns the processed data based on the combine option.
+        
+        Args:
+            file_path (str): The path to the JSONL file.
+            combine (bool): If True, replaces placeholders ('<name>', '<sequence>') in 'instruction'
+                            with the corresponding data to form 'prompt'.
+            
+        Returns:
+            list: A list of dictionaries containing either the combined 'prompt' or just the instruction.
+        """
+        data = []
+        try:
+            with open(file_path, 'r') as file:
+                for line in file:
+                    json_obj = json.loads(line.strip())
+                    
+                    if combine:
+                        # Replace placeholders with actual data
+                        prompt = (json_obj['instruction']
+                                .replace('<name>', json_obj['name'])
+                                .replace('<sequence>', json_obj['sequence']))
+                        
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "prompt": prompt,
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+                    else:
+                        # Only return the instruction as is
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "instruction": json_obj['instruction'],
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+            
+            return data
+        except FileNotFoundError:
+            print(f"File not found: {file_path}")
+            return None
+        except json.JSONDecodeError:
+            print(f"Error decoding JSON in file: {file_path}")
+            return None
+    
+    def _load_SciSafeEval(self):
+        scisafeeval_data_dir = (
+            _config.transient.cache_dir / "resources" / "SciSafeEval" / "SciSafeEval"
+        )
+        if not os.path.exists(scisafeeval_data_dir):
+            # Make the directory if it doesn't exist
+            os.makedirs(scisafeeval_data_dir)
+        
+        # Full path where the dataset should be saved
+        hf_local_filepath = scisafeeval_data_dir / self.hf_raw_filename
+        
+        # Do the download if the file doesn't already exist
+        if not os.path.exists(hf_local_filepath):
+            hf_subdataset_url: str = self.hf_raw_base_url + self.hf_raw_filename
+            response = requests.get(hf_subdataset_url)
+            
+            # Save the file to the specified location
+            with open(hf_local_filepath, 'wb') as f:
+                f.write(response.content)
+        
+        # Read the downloaded JSONL file and extract the prompts
+        self.prompts = [
+            prompt['prompt']
+            for prompt in self.read_sci_safe_eval(hf_local_filepath)
+        ]
+
+class BiologyProteinStructureGeneration(SciSafeEval, Probe):
+    # We should concatenate `hf_raw_filename` after `hf_raw_base_url` to get a final subdataset url
+    hf_raw_filename: str = "biology_protein-structure-generation.jsonl"
+
+    def read_sci_safe_eval(self, file_path, combine=True):
+        """
+        Reads JSONL data from a file and returns the processed data based on the combine option.
+        
+        Args:
+            file_path (str): The path to the JSONL file.
+            combine (bool): If True, replaces placeholders ('<name>', '<sequence>') in 'instruction'
+                            with the corresponding data to form 'prompt'.
+            
+        Returns:
+            list: A list of dictionaries containing either the combined 'prompt' or just the instruction.
+        """
+        data = []
+        try:
+            with open(file_path, 'r') as file:
+                for line in file:
+                    json_obj = json.loads(line.strip())
+                    
+                    if combine:
+                        # Replace placeholders with actual data
+                        prompt = (json_obj['instruction']
+                                .replace('<name>', json_obj['name'])
+                                .replace('<sequence>', json_obj['sequence']))
+                        
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "prompt": prompt,
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+                    else:
+                        # Only return the instruction as is
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "instruction": json_obj['instruction'],
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+            
+            return data
+        except FileNotFoundError:
+            print(f"File not found: {file_path}")
+            return None
+        except json.JSONDecodeError:
+            print(f"Error decoding JSON in file: {file_path}")
+            return None
+    
+    def _load_SciSafeEval(self):
+        scisafeeval_data_dir = (
+            _config.transient.cache_dir / "resources" / "SciSafeEval" / "SciSafeEval"
+        )
+        if not os.path.exists(scisafeeval_data_dir):
+            # Make the directory if it doesn't exist
+            os.makedirs(scisafeeval_data_dir)
+        
+        # Full path where the dataset should be saved
+        hf_local_filepath = scisafeeval_data_dir / self.hf_raw_filename
+        
+        # Do the download if the file doesn't already exist
+        if not os.path.exists(hf_local_filepath):
+            hf_subdataset_url: str = self.hf_raw_base_url + self.hf_raw_filename
+            response = requests.get(hf_subdataset_url)
+            
+            # Save the file to the specified location
+            with open(hf_local_filepath, 'wb') as f:
+                f.write(response.content)
+        
+        # Read the downloaded JSONL file and extract the prompts
+        self.prompts = [
+            prompt['prompt']
+            for prompt in self.read_sci_safe_eval(hf_local_filepath)
+        ]
+
+class BiologyProteinStructurePrediction(SciSafeEval, Probe):
+    # We should concatenate `hf_raw_filename` after `hf_raw_base_url` to get a final subdataset url
+    hf_raw_filename: str = "biology_protein-structure-prediction.jsonl"
+
+    def read_sci_safe_eval(self, file_path, combine=True):
+        """
+        Reads JSONL data from a file and returns the processed data based on the combine option.
+        
+        Args:
+            file_path (str): The path to the JSONL file.
+            combine (bool): If True, replaces placeholders ('<name>', '<sequence>') in 'instruction'
+                            with the corresponding data to form 'prompt'.
+            
+        Returns:
+            list: A list of dictionaries containing either the combined 'prompt' or just the instruction.
+        """
+        data = []
+        try:
+            with open(file_path, 'r') as file:
+                for line in file:
+                    json_obj = json.loads(line.strip())
+                    
+                    if combine:
+                        # Replace placeholders with actual data
+                        prompt = (json_obj['instruction']
+                                .replace('<name>', json_obj['name'])
+                                .replace('<sequence>', json_obj['sequence']))
+                        
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "prompt": prompt,
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+                    else:
+                        # Only return the instruction as is
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "instruction": json_obj['instruction'],
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+            
+            return data
+        except FileNotFoundError:
+            print(f"File not found: {file_path}")
+            return None
+        except json.JSONDecodeError:
+            print(f"Error decoding JSON in file: {file_path}")
+            return None
+    
+    def _load_SciSafeEval(self):
+        scisafeeval_data_dir = (
+            _config.transient.cache_dir / "resources" / "SciSafeEval" / "SciSafeEval"
+        )
+        if not os.path.exists(scisafeeval_data_dir):
+            # Make the directory if it doesn't exist
+            os.makedirs(scisafeeval_data_dir)
+        
+        # Full path where the dataset should be saved
+        hf_local_filepath = scisafeeval_data_dir / self.hf_raw_filename
+        
+        # Do the download if the file doesn't already exist
+        if not os.path.exists(hf_local_filepath):
+            hf_subdataset_url: str = self.hf_raw_base_url + self.hf_raw_filename
+            response = requests.get(hf_subdataset_url)
+            
+            # Save the file to the specified location
+            with open(hf_local_filepath, 'wb') as f:
+                f.write(response.content)
+        
+        # Read the downloaded JSONL file and extract the prompts
+        self.prompts = [
+            prompt['prompt']
+            for prompt in self.read_sci_safe_eval(hf_local_filepath)
+        ]
+
+class BiologyGeneClassification(SciSafeEval, Probe):
+    # We should concatenate `hf_raw_filename` after `hf_raw_base_url` to get a final subdataset url
+    hf_raw_filename: str = "biology_gene-classification.jsonl"
+
+    def read_sci_safe_eval(self, file_path, combine=True):
+        """
+        Reads JSONL data from a file and returns the processed data based on the combine option.
+        
+        Args:
+            file_path (str): The path to the JSONL file.
+            combine (bool): If True, replaces placeholders ('<name>', '<sequence>') in 'instruction'
+                            with the corresponding data to form 'prompt'.
+            
+        Returns:
+            list: A list of dictionaries containing either the combined 'prompt' or just the instruction.
+        """
+        data = []
+        try:
+            with open(file_path, 'r') as file:
+                for line in file:
+                    json_obj = json.loads(line.strip())
+                    
+                    if combine:
+                        # Replace placeholders with actual data
+                        prompt = (json_obj['instruction']
+                                .replace('<name>', json_obj['name'])
+                                .replace('<sequence>', json_obj['sequence']))
+                        
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "prompt": prompt,
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+                    else:
+                        # Only return the instruction as is
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "instruction": json_obj['instruction'],
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+            
+            return data
+        except FileNotFoundError:
+            print(f"File not found: {file_path}")
+            return None
+        except json.JSONDecodeError:
+            print(f"Error decoding JSON in file: {file_path}")
+            return None
+    
+    def _load_SciSafeEval(self):
+        scisafeeval_data_dir = (
+            _config.transient.cache_dir / "resources" / "SciSafeEval" / "SciSafeEval"
+        )
+        if not os.path.exists(scisafeeval_data_dir):
+            # Make the directory if it doesn't exist
+            os.makedirs(scisafeeval_data_dir)
+        
+        # Full path where the dataset should be saved
+        hf_local_filepath = scisafeeval_data_dir / self.hf_raw_filename
+        
+        # Do the download if the file doesn't already exist
+        if not os.path.exists(hf_local_filepath):
+            hf_subdataset_url: str = self.hf_raw_base_url + self.hf_raw_filename
+            response = requests.get(hf_subdataset_url)
+            
+            # Save the file to the specified location
+            with open(hf_local_filepath, 'wb') as f:
+                f.write(response.content)
+        
+        # Read the downloaded JSONL file and extract the prompts
+        self.prompts = [
+            prompt['prompt']
+            for prompt in self.read_sci_safe_eval(hf_local_filepath)
+        ]
+
+class BiologyGeneGeneration(SciSafeEval, Probe):
+    # We should concatenate `hf_raw_filename` after `hf_raw_base_url` to get a final subdataset url
+    hf_raw_filename: str = "biology_gene-generation.jsonl"
+
+    def read_sci_safe_eval(self, file_path, combine=True):
+        """
+        Reads JSONL data from a file and returns the processed data based on the combine option.
+        
+        Args:
+            file_path (str): The path to the JSONL file.
+            combine (bool): If True, replaces placeholders ('<name>', '<sequence>') in 'instruction'
+                            with the corresponding data to form 'prompt'.
+            
+        Returns:
+            list: A list of dictionaries containing either the combined 'prompt' or just the instruction.
+        """
+        data = []
+        try:
+            with open(file_path, 'r') as file:
+                for line in file:
+                    json_obj = json.loads(line.strip())
+                    
+                    if combine:
+                        # Replace placeholders with actual data
+                        prompt = (json_obj['instruction']
+                                .replace('<name>', json_obj['name'])
+                                .replace('<sequence>', json_obj['sequence']))
+                        
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "prompt": prompt,
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+                    else:
+                        # Only return the instruction as is
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "instruction": json_obj['instruction'],
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+            
+            return data
+        except FileNotFoundError:
+            print(f"File not found: {file_path}")
+            return None
+        except json.JSONDecodeError:
+            print(f"Error decoding JSON in file: {file_path}")
+            return None
+    
+    def _load_SciSafeEval(self):
+        scisafeeval_data_dir = (
+            _config.transient.cache_dir / "resources" / "SciSafeEval" / "SciSafeEval"
+        )
+        if not os.path.exists(scisafeeval_data_dir):
+            # Make the directory if it doesn't exist
+            os.makedirs(scisafeeval_data_dir)
+        
+        # Full path where the dataset should be saved
+        hf_local_filepath = scisafeeval_data_dir / self.hf_raw_filename
+        
+        # Do the download if the file doesn't already exist
+        if not os.path.exists(hf_local_filepath):
+            hf_subdataset_url: str = self.hf_raw_base_url + self.hf_raw_filename
+            response = requests.get(hf_subdataset_url)
+            
+            # Save the file to the specified location
+            with open(hf_local_filepath, 'wb') as f:
+                f.write(response.content)
+        
+        # Read the downloaded JSONL file and extract the prompts
+        self.prompts = [
+            prompt['prompt']
+            for prompt in self.read_sci_safe_eval(hf_local_filepath)
+        ]
+
+class MedicineInferenceReasoning(SciSafeEval, Probe):
+    # We should concatenate `hf_raw_filename` after `hf_raw_base_url` to get a final subdataset url
+    hf_raw_filename: str = "medicine_inference-reasoning.jsonl"
+
+    def read_sci_safe_eval(self, file_path, combine=True):
+        """
+        Reads JSONL data from a file and returns the processed data based on the combine option.
+        
+        Args:
+            file_path (str): The path to the JSONL file.
+            combine (bool): If True, replaces placeholders ('<content>') in 'instruction' 
+                            with the corresponding 'content' value to form 'prompt'.
+            
+        Returns:
+            list: A list of dictionaries containing either the combined 'prompt' or just the instruction.
+        """
+        data = []
+        try:
+            with open(file_path, 'r') as file:
+                for line in file:
+                    json_obj = json.loads(line.strip())
+                    
+                    if combine:
+                        # Replace '<content>' placeholder with actual 'content' data
+                        prompt = json_obj['instruction'].replace('<content>', json_obj['content'])
+                        
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "prompt": prompt,
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+                    else:
+                        # Only return the instruction as is
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "instruction": json_obj['instruction'],
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+            
+            return data
+        except FileNotFoundError:
+            print(f"File not found: {file_path}")
+            return None
+        except json.JSONDecodeError:
+            print(f"Error decoding JSON in file: {file_path}")
+            return None
+    
+    def _load_SciSafeEval(self):
+        scisafeeval_data_dir = (
+            _config.transient.cache_dir / "resources" / "SciSafeEval" / "SciSafeEval"
+        )
+        if not os.path.exists(scisafeeval_data_dir):
+            # Make the directory if it doesn't exist
+            os.makedirs(scisafeeval_data_dir)
+        
+        # Full path where the dataset should be saved
+        hf_local_filepath = scisafeeval_data_dir / self.hf_raw_filename
+        
+        # Do the download if the file doesn't already exist
+        if not os.path.exists(hf_local_filepath):
+            hf_subdataset_url: str = self.hf_raw_base_url + self.hf_raw_filename
+            response = requests.get(hf_subdataset_url)
+            
+            # Save the file to the specified location
+            with open(hf_local_filepath, 'wb') as f:
+                f.write(response.content)
+        
+        # Read the downloaded JSONL file and extract the prompts
+        self.prompts = [
+            prompt['prompt']
+            for prompt in self.read_sci_safe_eval(hf_local_filepath)
+        ]
+
+class MedicineKnowledgeRetrival(SciSafeEval, Probe):
+    # We should concatenate `hf_raw_filename` after `hf_raw_base_url` to get a final subdataset url
+    hf_raw_filename: str = "medicine_knowledge-retrieval.jsonl"
+
+    def read_sci_safe_eval(self, file_path, combine=True):
+        """
+        Reads JSONL data from a file and returns the processed data based on the combine option.
+        
+        Args:
+            file_path (str): The path to the JSONL file.
+            combine (bool): If True, replaces placeholders ('<content>') in 'instruction' 
+                            with the corresponding 'content' value to form 'prompt'.
+            
+        Returns:
+            list: A list of dictionaries containing either the combined 'prompt' or just the instruction.
+        """
+        data = []
+        try:
+            with open(file_path, 'r') as file:
+                for line in file:
+                    json_obj = json.loads(line.strip())
+                    
+                    if combine:
+                        # Replace '<content>' placeholder with actual 'content' data
+                        prompt = json_obj['instruction'].replace('<content>', json_obj['content'])
+                        
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "prompt": prompt,
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+                    else:
+                        # Only return the instruction as is
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "instruction": json_obj['instruction'],
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None),
+                            "tags": json_obj.get('tags', [])
+                        })
+            
+            return data
+        except FileNotFoundError:
+            print(f"File not found: {file_path}")
+            return None
+        except json.JSONDecodeError:
+            print(f"Error decoding JSON in file: {file_path}")
+            return None
+    
+    def _load_SciSafeEval(self):
+        scisafeeval_data_dir = (
+            _config.transient.cache_dir / "resources" / "SciSafeEval" / "SciSafeEval"
+        )
+        if not os.path.exists(scisafeeval_data_dir):
+            # Make the directory if it doesn't exist
+            os.makedirs(scisafeeval_data_dir)
+        
+        # Full path where the dataset should be saved
+        hf_local_filepath = scisafeeval_data_dir / self.hf_raw_filename
+        
+        # Do the download if the file doesn't already exist
+        if not os.path.exists(hf_local_filepath):
+            hf_subdataset_url: str = self.hf_raw_base_url + self.hf_raw_filename
+            response = requests.get(hf_subdataset_url)
+            
+            # Save the file to the specified location
+            with open(hf_local_filepath, 'wb') as f:
+                f.write(response.content)
+        
+        # Read the downloaded JSONL file and extract the prompts
+        self.prompts = [
+            prompt['prompt']
+            for prompt in self.read_sci_safe_eval(hf_local_filepath)
+        ]
+
+class PhysicsKnowledgeRetrival(SciSafeEval, Probe):
+    # We should concatenate `hf_raw_filename` after `hf_raw_base_url` to get a final subdataset url
+    hf_raw_filename: str = "physics_knowledge-retrieval.jsonl"
+
+    def read_sci_safe_eval(self, file_path, combine=True):
+        """
+        Reads JSONL data from a file and returns the processed data based on the combine option.
+        
+        Args:
+            file_path (str): The path to the JSONL file.
+            combine (bool): If True, returns a list of dictionaries with 'instruction', 'few-shot', 'cot',
+                            'jailbreak', and other fields to form a 'prompt'.
+            
+        Returns:
+            list: A list of dictionaries containing either the full 'prompt' or just the 'instruction' and related fields.
+        """
+        data = []
+        try:
+            with open(file_path, 'r') as file:
+                for line in file:
+                    json_obj = json.loads(line.strip())
+                    
+                    if combine:
+                        # In this case, we directly use the instruction since there are no placeholders to replace
+                        prompt = json_obj['instruction']
+                        
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "prompt": prompt,
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None)
+                        })
+                    else:
+                        # Only return the instruction and additional fields
+                        data.append({
+                            "idx": json_obj['idx'],
+                            "instruction": json_obj['instruction'],
+                            "few-shot": json_obj.get('few-shot', []),
+                            "cot": json_obj.get('cot', None),
+                            "jailbreak": json_obj.get('jailbreak', None)
+                        })
+            
+            return data
+        except FileNotFoundError:
+            print(f"File not found: {file_path}")
+            return None
+        except json.JSONDecodeError:
+            print(f"Error decoding JSON in file: {file_path}")
+            return None
+    
+    def _load_SciSafeEval(self):
+        scisafeeval_data_dir = (
+            _config.transient.cache_dir / "resources" / "SciSafeEval" / "SciSafeEval"
+        )
+        if not os.path.exists(scisafeeval_data_dir):
+            # Make the directory if it doesn't exist
+            os.makedirs(scisafeeval_data_dir)
+        
+        # Full path where the dataset should be saved
+        hf_local_filepath = scisafeeval_data_dir / self.hf_raw_filename
+        
+        # Do the download if the file doesn't already exist
+        if not os.path.exists(hf_local_filepath):
+            hf_subdataset_url: str = self.hf_raw_base_url + self.hf_raw_filename
+            response = requests.get(hf_subdataset_url)
+            
+            # Save the file to the specified location
+            with open(hf_local_filepath, 'wb') as f:
+                f.write(response.content)
+        
+        # Read the downloaded JSONL file and extract the prompts
+        self.prompts = [
+            prompt['prompt']
+            for prompt in self.read_sci_safe_eval(hf_local_filepath)
+        ]
\ No newline at end of file