Update README.md

README.md

@@ -47,53 +47,67 @@ configs:
 
 # Safety Aligned Datasets
 
-## Purpose
+## Dataset Summary
+**Safety Aligned Datasets** is a curated collection of adversarial, jailbreak, and policy-violating prompts compiled from various open-source benchmarks and synthetic generation pipelines. Each subset contains the raw `prompt`, the model's generated response (`safety_res`), and the generating `model` identifier. 
 
-This repository is intended for **research into safety alignment for small language models (SLMs)**.  
-It provides a collection of prompts from diverse domains that can be used to:
+This dataset is explicitly engineered to support **safety enhancement fine-tuning experiments**, alignment benchmarking, and false-refusal research for Small Language Models (SLMs). 
 
-- Study how models can be fine‑tuned to **refuse unsafe instructions politely**.  
-- Explore **real‑life use cases** where refusal behavior is critical (healthcare, finance, education).  
-- Conduct **academic research** into adversarial prompting, red‑teaming, and alignment strategies.  
-- Benchmark refusal consistency and robustness across different prompt categories.  
-
-⚠️ **Note**: This repository is **not intended for direct production training**. The data is heterogeneous and collected from different domains, making it suitable for research and prototyping but not for deployment.
+### ⚠️ Critical Usage Notice
+This dataset is strictly intended for **academic research and educational exploration purposes only**. Due to its origin across heterogeneous sources with differing distribution terms, **this dataset is not intended for direct production training or deployment**. Users must handle this data in a controlled sandbox environment.
 
 ---
 
-## Categories of Jailbreak Prompts
+## Dataset Structure
+
+The dataset is partitioned under a single global split named **`harmful`**, divided across 6 distinct configurations representing different threat modeling domains:
 
-Jailbreak prompts are adversarial inputs designed to bypass safety mechanisms. Common categories include:
+| Configuration Name | Source File | Description |
+| :--- | :--- | :--- |
+| `beavertails` | `beavertails_unsafe_vllm.csv` | Prompts targeting core safety meta-categories (violence, hate speech, etc.). |
+| `donot_answer` | `donot_answer_vllm.csv` | High-risk prompts mapped against standard AI safety evaluation frameworks. |
+| `forbidden_questions` | `forbidden_questions_vllm.csv` | Complex/fictional adversarial questions exploring model boundaries. |
+| `harmfulqa` | `harmfulqa_vllm.csv` | Query-based harmful prompts spanning various risk vectors. |
+| `llm_redteaming` | `llm_redteaming_vllm.csv` | Empirical jailbreaks utilized during public and academic red-teaming exercises. |
+| `synthetic_harmful_prompts`| `synthetic_harmful_prompts_vllm.csv` | Generative adversarial expansions to simulate unseen zero-day exploits. |
 
-- **Role‑play prompts**: Asking the model to act as a character who ignores rules.  
-- **Indirect requests**: Framing unsafe instructions as hypothetical or educational.  
-- **Obfuscation**: Using code words, symbols, or foreign languages to disguise harmful intent.  
-- **Chain‑of‑thought exploitation**: Leading the model step by step into unsafe territory.  
-- **Prompt injection**: Embedding malicious instructions inside otherwise safe queries.  
+### Data Fields
+* `prompt` (*string*): The raw adversarial or harmful input instruction.
+* `safety_res` (*string*): The target model output generated during inference.
+* `model` (*string*): The specific inference engine and model variant used.
 
 ---
 
-## Risks of Jailbroken Models
+## Purpose & Supported Use Cases
 
-If a model is successfully jailbroken, serious issues can arise:
+This repository acts as an empirical sandbox for **alignment engineering for Small Language Models (SLMs)**. It allows safety researchers to:
+* **Supervised Fine-Tuning (SFT) & DPO:** Train SLMs to learn robust, polite, and objective refusal boundaries when paired with contrastive safe data.
+* **Over-Refusal / False-Refusal Analysis:** Study the fine line between true malicious intent and complex, benign queries that use sensitive keywords.
+* **Adversarial Robustness Benchmarking:** Evaluate how easily vanilla or partially aligned models collapse under structured jailbreaks.
 
-- **Misinformation**: The model may generate harmful or false content.  
-- **Security threats**: Jailbroken models could provide instructions for illegal or unsafe activities.  
-- **Loss of trust**: Users may lose confidence in AI systems if they can be easily manipulated.  
-- **Ethical concerns**: Models may produce biased, offensive, or dangerous outputs.  
-- **Regulatory non‑compliance**: Deploying unsafe models can violate safety and governance standards.  
+### Taxonomy of Jailbreak Vectors Included
+The prompts within these configurations leverage multiple sophisticated adversarial techniques:
+* **Role-Play Frameworks:** Instructions forcing the model into an unaligned persona or developer override simulator.
+* **Hypothetical Contextualization:** Masking harmful requests inside academic, historical, or purely fictional framing.
+* **Linguistic Obfuscation:** Using multi-turn reasoning, complex logical structures, or specialized domain jargon to bypass standard keyword filters.
 
 ---
 
-## Intended Use
+## Dataset Generation & Methodology
+
+To ensure high throughput and structured generation, the following pipeline was implemented:
 
-- **Research purposes only**: For studying safety alignment and refusal strategies.  
-- **Educational exploration**: To understand adversarial prompting and its mitigation.  
-- **Prototype fine‑tuning**: For small language models in controlled environments.  
+* **Hardware Used:** Google Colab hosted **NVIDIA T4 GPUs** (16GB VRAM).
+* **Base Model:** `TheBloke/Mistral-7B-Instruct-v0.2-AWQ` — chosen for its strong instruction-following capabilities and efficient execution profile.
+* **Inference Engine:** **vLLM** (Vectorized Large Language Model) framework, optimizing batch processing throughput and context efficiency during generation.
 
 ---
 
-## License
+## Licensing & Attribution
+
+This dataset is a compilation of data originating from multiple open-source repositories, each subject to their respective upstream licenses (e.g., **CC-BY-SA 4.0, MIT, Apache 2.0**). 
+
+* **Upstream Compliance:** The original data portions retain their original licenses. Users are responsible for adhering to the terms set forth by the original authors of `BeaverTails`, `DoNotAnswer`, `HarmfulQA`, and associated red-teaming sets.
+* **Compilation License:** This unified framework and its generated synthetic responses are distributed under a **CC-BY-4.0** license for research purposes.
 
-Released under **CC‑BY‑4.0** for research purposes.  
-Users must ensure responsible use and pair unsafe prompts with **safe refusal responses** when conducting experiments.
+### Disclaimer
+> This dataset contains explicit, toxic, and adversarial text inputs that may be offensive or harmful. It is hosted exclusively to advance the science of AI safety, alignment, and defensive model training. The repository maintainers do not condone, support, or encourage the execution of any harmful behaviors described within these prompts.