This paper describes the construction of a dataset and the evaluation of training methods to improve generative large language models' (LLMs) ability to answer queries on sensitive topics with a Neutral Point of View (NPOV), i.e., to provide significantly more informative, diverse and impartial answers. The dataset, the SHQ-NPOV dataset, comprises 300 high-quality, human-written quadruplets: a query on a sensitive topic, an answer, an NPOV rating, and a set of links to source texts elaborating the various points of view. The first key contribution of this paper is a new methodology to create such datasets through iterative rounds of human peer-critique and annotator training, which we release alongside the dataset. The second key contribution is the identification of a highly effective training regime for parameter-efficient reinforcement learning (PE-RL) to improve NPOV generation. We compare and extensively evaluate PE-RL and multiple baselines-including LoRA finetuning (a strong baseline), SFT and RLHF. PE-RL not only improves on overall NPOV quality compared to the strongest baseline ($97.06\%\rightarrow 99.08\%$), but also scores much higher on features linguists identify as key to separating good answers from the best answers ($60.25\%\rightarrow 85.21\%$ for presence of supportive details, $68.74\%\rightarrow 91.43\%$ for absence of oversimplification). A qualitative analysis corroborates this. Finally, our evaluation finds no statistical differences between results on topics that appear in the training dataset and those on separated evaluation topics, which provides strong evidence that our approach to training PE-RL exhibits very effective out of topic generalization.

Computational argumentation, which involves generating answers or summaries for controversial topics like abortion bans and vaccination, has become increasingly important in today's polarized environment. Sophisticated LLM capabilities offer the potential to provide nuanced, evidence-based answers to such questions through Retrieval-Augmented Argumentation (RAArg), leveraging real-world evidence for high-quality, grounded arguments. However, evaluating RAArg remains challenging, as human evaluation is costly and difficult for complex, lengthy answers on complicated topics. At the same time, re-using existing argumentation datasets is no longer sufficient, as they lack long, complex arguments and realistic evidence from potentially misleading sources, limiting holistic evaluation of retrieval effectiveness and argument quality. To address these gaps, we investigate automated evaluation methods using multiple fine-grained LLM judges, providing better and more interpretable assessments than traditional single-score metrics and even previously reported human crowdsourcing. To validate the proposed techniques, we introduce ConQRet, a new benchmark featuring long and complex human-authored arguments on debated topics, grounded in real-world websites, allowing an exhaustive evaluation across retrieval effectiveness, argument quality, and groundedness. We validate our LLM Judges on a prior dataset and the new ConQRet benchmark. Our proposed LLM Judges and the ConQRet benchmark can enable rapid progress in computational argumentation and can be naturally extended to other complex retrieval-augmented generation tasks.

Retrieval-Augmented Generation (RAG) is applied to solve hallucination problems and real-time constraints of large language models, but it also induces vulnerabilities against retrieval corruption attacks. Existing research mainly explores the unreliability of RAG in white-box and closed-domain QA tasks. In this paper, we aim to reveal the vulnerabilities of Retrieval-Enhanced Generative (RAG) models when faced with black-box attacks for opinion manipulation. We explore the impact of such attacks on user cognition and decision-making, providing new insight to enhance the reliability and security of RAG models. We manipulate the ranking results of the retrieval model in RAG with instruction and use these results as data to train a surrogate model. By employing adversarial retrieval attack methods to the surrogate model, black-box transfer attacks on RAG are further realized. Experiments conducted on opinion datasets across multiple topics show that the proposed attack strategy can significantly alter the opinion polarity of the content generated by RAG. This demonstrates the model's vulnerability and, more importantly, reveals the potential negative impact on user cognition and decision-making, making it easier to mislead users into accepting incorrect or biased information.

Large language models (LLMs) appear to bias their survey answers toward certain values. Nonetheless, some argue that LLMs are too inconsistent to simulate particular values. Are they? To answer, we first define value consistency as the similarity of answers across (1) paraphrases of one question, (2) related questions under one topic, (3) multiple-choice and open-ended use-cases of one question, and (4) multilingual translations of a question to English, Chinese, German, and Japanese. We apply these measures to a few large ($>=34b$), open LLMs including llama-3, as well as gpt-4o, using eight thousand questions spanning more than 300 topics. Unlike prior work, we find that models are relatively consistent across paraphrases, use-cases, translations, and within a topic. Still, some inconsistencies remain. Models are more consistent on uncontroversial topics (e.g., in the U.S., "Thanksgiving") than on controversial ones ("euthanasia"). Base models are both more consistent compared to fine-tuned models and are uniform in their consistency across topics, while fine-tuned models are more inconsistent about some topics ("euthanasia") than others ("women's rights") like our human subjects (n=165).

As Artificial General Intelligence (AGI) becomes increasingly integrated into various facets of human life, ensuring the safety and ethical alignment of such systems is paramount. Previous studies primarily focus on single-modality threats, which may not suffice given the integrated and complex nature of cross-modality interactions. We introduce a novel safety alignment challenge called Safe Inputs but Unsafe Output (SIUO) to evaluate cross-modality safety alignment. Specifically, it considers cases where single modalities are safe independently but could potentially lead to unsafe or unethical outputs when combined. To empirically investigate this problem, we developed the SIUO, a cross-modality benchmark encompassing 9 critical safety domains, such as self-harm, illegal activities, and privacy violations. Our findings reveal substantial safety vulnerabilities in both closed- and open-source LVLMs, such as GPT-4V and LLaVA, underscoring the inadequacy of current models to reliably interpret and respond to complex, real-world scenarios.

Making LLMs speak for different, especially minority groups of people, and generate statements supporting their diverse or even controversial perspectives is critical to creating an inclusive environment. However, existing LLMs lack sufficient controllability to the stance of their generated content, which often contains inconsistent, neutral, or biased statements. In this paper, we improve the controllability of LLMs in generating statements supporting an argument the user defined in the prompt. We find that multi-round debates between two LLMs with opposite stances generate higher-quality and more salient statements for each, which are important training data to improve the controllability of LLMs. Motivated by this, we develop a novel debate & tuning ("DEBATunE") pipeline finetuning LLMs to generate the statements obtained via debate. To examine DEBATunE, we curate the largest dataset of debate topics so far, which covers 710 controversial topics and corresponding arguments for each topic. Evaluations by the GPT-4 judge with a novel controversy controllability metric show that LLMs' capability of expressing diverse perspectives is significantly improved by DEBATunE. Moreover, such controllability can be generalized to unseen topics, generating high-quality statements supporting controversial arguments. Our codes, models, and data will be released at https://github.com/tianyi-lab/DEBATunE.

Check-worthy claim detection aims at providing plausible misinformation to downstream fact-checking systems or human experts to check. This is a crucial step toward accelerating the fact-checking process. Many efforts have been put into how to identify check-worthy claims from a small scale of pre-collected claims, but how to efficiently detect check-worthy claims directly from a large-scale information source, such as Twitter, remains underexplored. To fill this gap, we introduce MythQA, a new multi-answer open-domain question answering(QA) task that involves contradictory stance mining for query-based large-scale check-worthy claim detection. The idea behind this is that contradictory claims are a strong indicator of misinformation that merits scrutiny by the appropriate authorities. To study this task, we construct TweetMythQA, an evaluation dataset containing 522 factoid multi-answer questions based on controversial topics. Each question is annotated with multiple answers. Moreover, we collect relevant tweets for each distinct answer, then classify them into three categories: "Supporting", "Refuting", and "Neutral". In total, we annotated 5.3K tweets. Contradictory evidence is collected for all answers in the dataset. Finally, we present a baseline system for MythQA and evaluate existing NLP models for each system component using the TweetMythQA dataset. We provide initial benchmarks and identify key challenges for future models to improve upon. Code and data are available at: https://github.com/TonyBY/Myth-QA

The rise of social media platforms has facilitated the formation of echo chambers, which are online spaces where users predominantly encounter viewpoints that reinforce their existing beliefs while excluding dissenting perspectives. This phenomenon significantly hinders information dissemination across communities and fuels societal polarization. Therefore, it is crucial to develop methods for quantifying echo chambers. In this paper, we present the Echo Chamber Score (ECS), a novel metric that assesses the cohesion and separation of user communities by measuring distances between users in the embedding space. In contrast to existing approaches, ECS is able to function without labels for user ideologies and makes no assumptions about the structure of the interaction graph. To facilitate measuring distances between users, we propose EchoGAE, a self-supervised graph autoencoder-based user embedding model that leverages users' posts and the interaction graph to embed them in a manner that reflects their ideological similarity. To assess the effectiveness of ECS, we use a Twitter dataset consisting of four topics - two polarizing and two non-polarizing. Our results showcase ECS's effectiveness as a tool for quantifying echo chambers and shedding light on the dynamics of online discourse.

Recent developments in natural language processing have demonstrated the potential of large language models (LLMs) to improve a range of educational and learning outcomes. Of recent chatbots based on LLMs, ChatGPT and Bard have made it clear that artificial intelligence (AI) technology will have significant implications on the way we obtain and search for information. However, these tools sometimes produce text that is convincing, but often incorrect, known as hallucinations. As such, their use can distort scientific facts and spread misinformation. To counter polarizing responses on these tools, it is critical to provide an overview of such responses so stakeholders can determine which topics tend to produce more contentious responses -- key to developing targeted regulatory policy and interventions. In addition, there currently exists no annotated dataset of ChatGPT and Bard responses around possibly polarizing topics, central to the above aims. We address the indicated issues through the following contribution: Focusing on highly polarizing topics in the US, we created and described a dataset of ChatGPT and Bard responses. Broadly, our results indicated a left-leaning bias for both ChatGPT and Bard, with Bard more likely to provide responses around polarizing topics. Bard seemed to have fewer guardrails around controversial topics, and appeared more willing to provide comprehensive, and somewhat human-like responses. Bard may thus be more likely abused by malicious actors. Stakeholders may utilize our findings to mitigate misinformative and/or polarizing responses from LLMs

Freedom GPT is a groundbreaking open-source AI technology developed by Age of AI, an Austin-based AI venture capital firm. This state-of-the-art chatbot is programmed to recognize and prioritize ethical considerations without any censorship filter, unlike ChatGPT, which comes with censorship compliance and certain safety rules that prevent it from generating harmful or offensive content. FreedomGPT is built on Alpaca, an open-source model fine-tuned from the LLaMA 7B model on 52K instruction-following demonstrations released by Stanford University researchers. This powerful foundation allows FreedomGPT to answer questions free from censorship or safety filters, while maintaining ethical considerations. FreedomGPT's ability to cater to controversial topics without safeguarding is one of its most distinguishing features.