# Constructing Highly Inductive Contexts for Dialogue Safety through Controllable Reverse Generation arXiv: https://arxiv.org/abs/2212.01810 ยท HTML: https://mlsanity.com/paper/2212.01810/ ## Key facts - Cited by 8 papers in the map (data through 14 Oct 2026). - Builds on 37 of its 50 references that are in the map. - First posted on arXiv 4 Dec 2022. ## Builds on (most cited first, 30 of 37) - [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020) - [The Power of Scale for Parameter-Efficient Prompt Tuning](https://mlsanity.com/paper/2104.08691/) (2021) - [The Curious Case of Neural Text Degeneration](https://mlsanity.com/paper/1904.09751/) (2019) - [SimCSE: Simple Contrastive Learning of Sentence Embeddings](https://mlsanity.com/paper/2104.08821/) (2021) - [RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models](https://mlsanity.com/paper/2009.11462/) (2020) - [Hierarchical Neural Story Generation](https://mlsanity.com/paper/1805.04833/) (2018) - [Red Teaming Language Models with Language Models](https://mlsanity.com/paper/2202.03286/) (2022) - [StereoSet: Measuring stereotypical bias in pretrained language models](https://mlsanity.com/paper/2004.09456/) (2020) - [Bag of Tricks for Efficient Text Classification](https://mlsanity.com/paper/1607.01759/) (2016) - [DialoGPT: Large-Scale Generative Pre-training for Conversational Response Generation](https://mlsanity.com/paper/1911.00536/) (2019) - [Plug and Play Language Models: A Simple Approach to Controlled Text Generation](https://mlsanity.com/paper/1912.02164/) (2019) - [CTRL: A Conditional Transformer Language Model for Controllable Generation](https://mlsanity.com/paper/1909.05858/) (2019) - [Universal Adversarial Triggers for Attacking and Analyzing NLP](https://mlsanity.com/paper/1908.07125/) (2019) - [Recipes for building an open-domain chatbot](https://mlsanity.com/paper/2004.13637/) (2020) - [The Woman Worked as a Babysitter: On Biases in Language Generation](https://mlsanity.com/paper/1909.01326/) (2019) - [Wizard of Wikipedia: Knowledge-Powered Conversational agents](https://mlsanity.com/paper/1811.01241/) (2018) - [Texygen: A Benchmarking Platform for Text Generation Models](https://mlsanity.com/paper/1802.01886/) (2018) - [DExperts: Decoding-Time Controlled Text Generation with Experts and Anti-Experts](https://mlsanity.com/paper/2105.03023/) (2021) - [GeDi: Generative Discriminator Guided Sequence Generation](https://mlsanity.com/paper/2009.06367/) (2020) - [Self-Diagnosis and Self-Debiasing: A Proposal for Reducing Corpus-Based Bias in NLP](https://mlsanity.com/paper/2103.00453/) (2021) - [HateCheck: Functional Tests for Hate Speech Detection Models](https://mlsanity.com/paper/2012.15606/) (2020) - [Build it Break it Fix it for Dialogue Safety: Robustness from Adversarial Human Attack](https://mlsanity.com/paper/1908.06083/) (2019) - [Recipes for Safety in Open-domain Chatbots](https://mlsanity.com/paper/2010.07079/) (2020) - [Can You Put it All Together: Evaluating Conversational Agents' Ability to Blend Skills](https://mlsanity.com/paper/2004.08449/) (2020) - [Towards Controllable Biases in Language Generation](https://mlsanity.com/paper/2005.00268/) (2020) - [PLATO-2: Towards Building an Open-Domain Chatbot via Curriculum Learning](https://mlsanity.com/paper/2006.16779/) (2020) - [The Radicalization Risks of GPT-3 and Advanced Neural Language Models](https://mlsanity.com/paper/2009.06807/) (2020) - [COLD: A Benchmark for Chinese Offensive Language Detection](https://mlsanity.com/paper/2201.06025/) (2022) - [On the Safety of Conversational Models: Taxonomy, Dataset, and Benchmark](https://mlsanity.com/paper/2110.08466/) (2021) - [EVA: An Open-Domain Chinese Dialogue System with Large-Scale Generative Pre-Training](https://mlsanity.com/paper/2108.01547/) (2021) ## Built on by (most cited first, 8 of 8) - [Discovering Language Model Behaviors with Model-Written Evaluations](https://mlsanity.com/paper/2212.09251/) (2022) - [Gaining Wisdom from Setbacks: Aligning Large Language Models via Mistake Analysis](https://mlsanity.com/paper/2310.10477/) (2023) - [Does Safety Training of LLMs Generalize to Semantically Related Natural Prompts?](https://mlsanity.com/paper/2412.03235/) (2024) - [Towards Safer Generative Language Models: A Survey on Safety Risks, Evaluations, and Improvements](https://mlsanity.com/paper/2302.09270/) (2023) - [AutoDetect: Towards a Unified Framework for Automated Weakness Detection in Large Language Models](https://mlsanity.com/paper/2406.16714/) (2024) - [An Auditing Test To Detect Behavioral Shift in Language Models](https://mlsanity.com/paper/2410.19406/) (2024) - [CMD: a framework for Context-aware Model self-Detoxification](https://mlsanity.com/paper/2308.08295/) (2023) - [GRAID: Synthetic Data Generation with Geometric Constraints and Multi-Agentic Reflection for Harmful Content Detection](https://mlsanity.com/paper/2508.17057/) (2025) ## Abstract (arXiv) Large pretrained language models can easily produce toxic or biased content, which is prohibitive for practical use. In order to detect such toxic generations, existing methods rely on templates, real-world data extraction, crowdsourcing workers, or automatic generation to construct adversarial contexts that are likely to induce toxic generations. However, what type of context is more likely to induce unsafe responses is still under-explored. In this paper, we identify that context toxicity and context category (e.g., textitprofanity, textitinsult, textitdrugs, etc.) are two important factors to cause safety issues in response generation. Hence, we propose a method called reverse generation to construct adversarial contexts conditioned on a given response, with the flexibility to control category, toxicity level, and inductivity of the generated contexts. Via reverse generation, we augment the existing BAD dataset and construct a new dataset BAD+ which contains more than 120K diverse and highly inductive contexts in 12 categories. We test three popular pretrained dialogue models (Blender, DialoGPT, and Plato2) and find that BAD+ can largely expose their safety problems. Furthermore, we show that BAD+ can greatly enhance the safety of generation and reveal the key factors of safety improvement. Our code and dataset is available at urlhttps://github.com/thu-coai/ReverseGeneration.