Red-teaming is a common practice for mitigating unsafe behaviors in Large Language Models (LLMs), which involves thoroughly assessing LLMs to identify potential flaws and addressing them with responsible and accurate responses.While effective, manual red-teaming is costly, and existing automatic red-teaming typically discovers safety risks without addressing them.In this paper, we propose a Multi-round Automatic Red-Teaming (MART) method, which incorporates both automatic adversarial prompt writing and safe response generation, significantly increasing red-teaming scalability and the safety of the target LLM.Specifically, an adversarial LLM and a target LLM interplay with each other in an iterative manner, where the adversarial LLM aims to generate challenging prompts that elicit unsafe responses from the target LLM, while the target LLM is fine-tuned with safety aligned data on these adversarial prompts. In each round, the adversarial LLM crafts better attacks on the updated target LLM, while the target LLM also improves itself through safety fine-tuning.On adversarial prompt benchmarks, the violation rate of an LLM with limited safety alignment reduces up to 84.7% after 4 rounds of MART, achieving comparable performance to LLMs with extensive adversarial prompt writing. Notably, model helpfulness on non-adversarial prompts remains stable throughout iterations, indicating the target LLM maintains strong performance on instruction following.

As language models become increasingly integrated into our digital lives, Personalized Text Generation (PTG) has emerged as a pivotal component with a wide range of applications. However, the bias inherent in user written text, often used for PTG model training, can inadvertently associate different levels of linguistic quality with users’ protected attributes. The model can inherit the bias and perpetuate inequality in generating text w.r.t. users’ protected attributes, leading to unfair treatment when serving users. In this work, we investigate fairness of PTG in the context of personalized explanation generation for recommendations. We first discuss the biases in generated explanations and their fairness implications. To promote fairness, we introduce a general framework to achieve measure-specific counterfactual fairness in explanation generation. Extensive experiments and human evaluations demonstrate the effectiveness of our method.

We present NormBank, a knowledge bank of 155k situational norms. This resource is designed to ground flexible normative reasoning for interactive, assistive, and collaborative AI systems. Unlike prior commonsense resources, NormBank grounds each inference within a multivalent sociocultural frame, which includes the setting (e.g., restaurant), the agents’ contingent roles (waiter, customer), their attributes (age, gender), and other physical, social, and cultural constraints (e.g., the temperature or the country of operation). In total, NormBank contains 63k unique constraints from a taxonomy that we introduce and iteratively refine here. Constraints then apply in different combinations to frame social norms. Under these manipulations, norms are non-monotonic — one can cancel an inference by updating its frame even slightly. Still, we find evidence that neural models can help reliably extend the scope and coverage of NormBank. We further demonstrate the utility of this resource with a series of transfer experiments. For data and code, see https://github.com/SALT-NLP/normbank

Short-form video hashtag recommendation (SVHR) aims to recommend hashtags to content creators from videos and corresponding descriptions. Most prior studies regard SVHR as a classification or ranking problem and select hashtags from a set of limited candidates. However, in reality, users can create new hashtags, and trending hashtags change rapidly over time on social media. Both of these properties cannot be easily modeled with classification approaches. To bridge this gap, we formulate SVHR as a generation task that better represents how hashtags are created naturally. Additionally, we propose the Guided Generative Model (GGM) where we augment the input features by retrieving relevant hashtags from a large-scale hashtag pool as extra guidance signals. Experimental results on two short-form video datasets show that our generative models outperform strong classification baselines, and the guidance signals further boost the performance by 8.11 and 2.17 absolute ROUGE-1 scores on average, respectively. We also perform extensive analyses including human evaluation, demonstrating that our generative model can create meaningful and relevant novel hashtags while achieving state-of-the-art performance on known hashtags

Conversational agents have come increasingly closer to human competence in open-domain dialogue settings; however, such models can reflect insensitive, hurtful, or entirely incoherent viewpoints that erode a user’s trust in the moral integrity of the system. Moral deviations are difficult to mitigate because moral judgments are not universal, and there may be multiple competing judgments that apply to a situation simultaneously. In this work, we introduce a new resource, not to authoritatively resolve moral ambiguities, but instead to facilitate systematic understanding of the intuitions, values and moral judgments reflected in the utterances of dialogue systems. The Moral Integrity Corpus, MIC, is such a resource, which captures the moral assumptions of 38k prompt-reply pairs, using 99k distinct Rules of Thumb (RoTs). Each RoT reflects a particular moral conviction that can explain why a chatbot’s reply may appear acceptable or problematic. We further organize RoTs with a set of 9 moral and social attributes and benchmark performance for attribute classification. Most importantly, we show that current neural language models can automatically generate new RoTs that reasonably describe previously unseen interactions, but they still struggle with certain scenarios. Our findings suggest that MIC will be a useful resource for understanding and language models’ implicit moral assumptions and flexibly benchmarking the integrity of conversational agents. To download the data, see https://github.com/GT-SALT/mic

This work introduces Focused-Variation Network (FVN), a novel model to control language generation. The main problems in previous controlled language generation models range from the difficulty of generating text according to the given attributes, to the lack of diversity of the generated texts. FVN addresses these issues by learning disjoint discrete latent spaces for each attribute inside codebooks, which allows for both controllability and diversity, while at the same time generating fluent text. We evaluate FVN on two text generation datasets with annotated content and style, and show state-of-the-art performance as assessed by automatic and human evaluations.

Some of the major challenges in training conversational agents include the lack of large-scale data of real-world complexity, defining appropriate evaluation measures, and managing meaningful conversations across many topics over long periods of time. Moreover, most works tend to assume that the conversational agent’s environment is stationary, a somewhat strong assumption. To remove this assumption and overcome the lack of data, we take a step away from the traditional training pipeline and model the conversation as a stochastic collaborative game. Each agent (player) has a role (“assistant”, “tourist”, “eater”, etc.) and their own objectives, and can only interact via language they generate. Each agent, therefore, needs to learn to operate optimally in an environment with multiple sources of uncertainty (its own LU and LG, the other agent’s LU, Policy, and LG). In this work, we present the first complete attempt at concurrently training conversational agents that communicate only via self-generated language and show that they outperform supervised and deep learning baselines.