Challenge: a new method for evaluating chatbot safety is proposed to mimic human-generated data . a bot-adversarial dialogue model learns undesirable features from this data, a study finds .
Approach: They propose a human-and-model-in-the-loop framework for evaluating toxicity of chatbots . they propose two methods for safe conversational agents by either training on data or ”baking-in” safety to the generative model itself.
Outcome: The proposed methods are safer than existing models while maintaining usability metrics, the authors say . they show that the proposed methods can be used to make safer models with human-model interactions .

Similar Papers

Using In-Context Learning to Improve Dialogue Safety (2023.findings-emnlp)

Copied to clipboard

Challenge: Recent work has highlighted safety issues with large neural-based conversational models.
Approach: They propose a retrieval-based approach for reducing bias and toxicity in chatbot responses . they retrieve demonstrations of safe responses to similar dialogue contexts to generate a response .
Outcome: The proposed method reduces bias and toxicity in three chatbot models . it can be used in compliment to existing dialogue safety approaches, such as RLHF.
On the Safety of Conversational Models: Taxonomy, Dataset, and Benchmark (2022.findings-acl)

Copied to clipboard

Challenge: Dialogue safety problems severely limit the real-world deployment of generative conversational models.
Approach: They propose a taxonomy for dialogue safety specifically designed to capture unsafe behaviors in human-bot dialogue settings.
Outcome: The proposed taxonomy captures unsafe behaviors in human-bot dialogue settings with rich context-sensitive unsafe examples.
Robust Conversational Agents against Imperceptible Toxicity Triggers (2022.naacl-main)

Copied to clipboard

Challenge: Existing work to generate adversarial attacks is costly and not scalable . despite the abundance of research in this area, little attention has been given to adversarials .
Approach: They propose an adversarial attack mechanism that mitigates toxic language generation . they propose a defense mechanism that is scalable and can be generalized .
Outcome: The proposed defense is effective at avoiding toxic language generation even against imperceptible toxicity triggers while preserving conversational flow.
ProsocialDialog: A Prosocial Backbone for Conversational Agents (2022.emnlp-main)

Copied to clipboard

Challenge: Existing dialogue systems fail to respond properly to potentially unsafe user utterances . existing systems either ignore or passively agree with unsafe content .
Approach: They introduce a dataset to teach conversational agents to respond to problematic content following social norms.
Outcome: The proposed dataset shows that ProsocialDialog generates more socially acceptable dialogues than existing models.
SafeConv: Explaining and Correcting Conversational Unsafe Behavior (2023.acl-long)

Copied to clipboard

Challenge: Existing datasets do not provide enough annotation to explain unsafe behavior . current chatbots generate toxic and offensive responses, which can be dangerous .
Approach: They construct a dataset called SafeConv that provides comprehensive annotations for chatbots . they compare safe alternatives to rewrite unsafe responses .
Outcome: The proposed model can explain unsafe behavior and detoxify chatbots, the authors show . the proposed model is able to detect unsafe utterances, extract unsafe spans, and convert unsafe responses to safe versions.
Build it Break it Fix it for Dialogue Safety: Robustness from Adversarial Human Attack (D19-1)

Copied to clipboard

Challenge: Detecting offensive language in the context of a dialogue is an increasingly important application of natural language processing.
Approach: They propose to train a model to be robust to such attacks by iterative build it, break it, fix it scheme with humans and models in the loop.
Outcome: The proposed model is significantly more robust to such human attacks than previous systems.
GrounDial: Human-norm Grounded Safe Dialog Response Generation (2024.findings-eacl)

Copied to clipboard

Challenge: Recent conversational AI systems generate unsafe responses agreeing to offensive user input or including toxic content.
Approach: They propose a method where response safety is achieved by grounding responses to commonsense social rules without fine-tuning.
Outcome: The proposed approach is quantitatively and qualitatively safer even without additional data or tuning.
Synthesizing Adversarial Negative Responses for Robust Response Ranking and Evaluation (2021.findings-acl)

Copied to clipboard

Challenge: Open-domain neural dialogue models have achieved high performance in response ranking and evaluation tasks.
Approach: They propose methods for automatically creating adversarial negative training data . they use mask-and-fill and keyword-guided approaches to generate negative examples .
Outcome: The proposed approaches outperform baseline models in providing informative negative examples for training dialogue systems.
Adversarial DPO: Harnessing Harmful Data for Reducing Toxicity with Minimal Impact on Coherence and Evasiveness in Dialogue Agents (2024.findings-naacl)

Copied to clipboard

Challenge: Existing toxicity within large language models can negatively impact the user experience, causing performance degradation.
Approach: They propose an adversarial DPO algorithm that improves direct preference optimization (DPO) by incorporating harmful data into the generative model.
Outcome: The proposed training algorithm improves the model’s resilience against harmful conversations while minimizing performance degradation.
Are Personalized Stochastic Parrots More Dangerous? Evaluating Persona Biases in Dialogue Systems (2023.findings-emnlp)

Copied to clipboard

Challenge: Recent advances in Large Language Models enable them to follow freeform instructions, including imitating generic or specific demographic personas in conversations.
Approach: They propose to investigate persona biases by experimenting with UNIVERSALPERSONA, a model that incorporates both generic and specific personas.
Outcome: The proposed model systematically measures persona biases in harmful expression and harmful agreement.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations