Intrinsic self-correct was a method that instructed large language models (LLMs) to verify and correct their responses without external feedback. Unfortunately, the study concluded that the LLMs could not self-correct reasoning yet. We find that a simple yet effective verification method can unleash inherent capabilities of the LLMs. That is to mask a key condition in the question, add the current response to construct a verification question, and predict the condition to verify the response. The condition can be an entity in an open-domain question or a numeric value in a math question, which requires minimal effort (via prompting) to identify. We propose an iterative verify-then-correct framework to progressively identify and correct (probably) false responses, named ProCo. We conduct experiments on three reasoning tasks. On average, ProCo, with GPT-3.5-Turbo as the backend LLM, yields $+6.8$ exact match on four open-domain question answering datasets, $+14.1$ accuracy on three arithmetic reasoning datasets, and $+9.6$ accuracy on a commonsense reasoning dataset, compared to Self-Correct.

Humans are prone to cognitive distortions -- biased thinking patterns that lead to exaggerated responses to specific stimuli, albeit in very different contexts. This paper demonstrates that advanced Multimodal Large Language Models (MLLMs) exhibit similar tendencies. While these models are designed to respond queries under safety mechanism, they sometimes reject harmless queries in the presence of certain visual stimuli, disregarding the benign nature of their contexts. As the initial step in investigating this behavior, we identify three types of stimuli that trigger the oversensitivity of existing MLLMs: Exaggerated Risk, Negated Harm, and Counterintuitive Interpretation. To systematically evaluate MLLMs' oversensitivity to these stimuli, we propose the Multimodal OverSenSitivity Benchmark (MOSSBench). This toolkit consists of 300 manually collected benign multimodal queries, cross-verified by third-party reviewers (AMT). Empirical studies using MOSSBench on 20 MLLMs reveal several insights: (1). Oversensitivity is prevalent among SOTA MLLMs, with refusal rates reaching up to 76% for harmless queries. (2). Safer models are more oversensitive: increasing safety may inadvertently raise caution and conservatism in the model's responses. (3). Different types of stimuli tend to cause errors at specific stages -- perception, intent reasoning, and safety judgement -- in the response process of MLLMs. These findings highlight the need for refined safety mechanisms that balance caution with contextually appropriate responses, improving the reliability of MLLMs in real-world applications. We make our project available at https://turningpoint-ai.github.io/MOSSBench/.

Machine learning (ML) models, demonstrably powerful, suffer from a lack of interpretability. The absence of transparency, often referred to as the black box nature of ML models, undermines trust and urges the need for efforts to enhance their explainability. Explainable AI (XAI) techniques address this challenge by providing frameworks and methods to explain the internal decision-making processes of these complex models. Techniques like Counterfactual Explanations (CF) and Feature Importance play a crucial role in achieving this goal. Furthermore, high-quality and diverse data remains the foundational element for robust and trustworthy ML applications. In many applications, the data used to train ML and XAI explainers contain sensitive information. In this context, numerous privacy-preserving techniques can be employed to safeguard sensitive information in the data, such as differential privacy. Subsequently, a conflict between XAI and privacy solutions emerges due to their opposing goals. Since XAI techniques provide reasoning for the model behavior, they reveal information relative to ML models, such as their decision boundaries, the values of features, or the gradients of deep learning models when explanations are exposed to a third entity. Attackers can initiate privacy breaching attacks using these explanations, to perform model extraction, inference, and membership attacks. This dilemma underscores the challenge of finding the right equilibrium between understanding ML decision-making and safeguarding privacy.

Multilingual LLMs often have knowledge disparities across languages, with larger gaps in under-resourced languages. Teaching LLMs to abstain in the face of knowledge gaps is thus a promising strategy to mitigate hallucinations in multilingual settings. However, previous studies on LLM abstention primarily focus on English; we find that directly applying existing solutions beyond English results in up to 20.5% performance gaps between high and low-resource languages, potentially due to LLMs' drop in calibration and reasoning beyond a few resource-rich languages. To this end, we propose strategies to enhance LLM abstention by learning from multilingual feedback, where LLMs self-reflect on proposed answers in one language by generating multiple feedback items in related languages: we show that this helps identifying the knowledge gaps across diverse languages, cultures, and communities. Extensive experiments demonstrate that our multilingual feedback approach outperforms various strong baselines, achieving up to 9.2% improvement for low-resource languages across three black-box and open models on three datasets, featuring open-book, closed-book, and commonsense QA. Further analysis reveals that multilingual feedback is both an effective and a more equitable abstain strategy to serve diverse language speakers, and cultural factors have great impact on language selection and LLM abstention behavior, highlighting future directions for multilingual and multi-cultural reliable language modeling.

Existing cross-document event coreference resolution models, which either compute mention similarity directly or enhance mention representation by extracting event arguments (such as location, time, agent, and patient), lacking the ability to utilize document-level information. As a result, they struggle to capture long-distance dependencies. This shortcoming leads to their underwhelming performance in determining coreference for the events where their argument information relies on long-distance dependencies. In light of these limitations, we propose the construction of document-level Rhetorical Structure Theory (RST) trees and cross-document Lexical Chains to model the structural and semantic information of documents. Subsequently, cross-document heterogeneous graphs are constructed and GAT is utilized to learn the representations of events. Finally, a pair scorer calculates the similarity between each pair of events and co-referred events can be recognized using standard clustering algorithm. Additionally, as the existing cross-document event coreference datasets are limited to English, we have developed a large-scale Chinese cross-document event coreference dataset to fill this gap, which comprises 53,066 event mentions and 4,476 clusters. After applying our model on the English and Chinese datasets respectively, it outperforms all baselines by large margins.

The location of knowledge within Generative Pre-trained Transformer (GPT)-like models has seen extensive recent investigation. However, much of the work is focused towards determining locations of individual facts, with the end goal being the editing of facts that are outdated, erroneous, or otherwise harmful, without the time and expense of retraining the entire model. In this work, we investigate a broader view of knowledge location, that of concepts or clusters of related information, instead of disparate individual facts. To do this, we first curate a novel dataset, called DARC, that includes a total of 34 concepts of ~120K factual statements divided into two types of hierarchical categories, namely taxonomy and meronomy. Next, we utilize existing causal mediation analysis methods developed for determining regions of importance for individual facts and apply them to a series of related categories to provide detailed investigation into whether concepts are associated with distinct regions within these models. We find that related categories exhibit similar areas of importance in contrast to less similar categories. However, fine-grained localization of individual category subsets to specific regions is not apparent.

Clustering is a fundamental problem in machine learning and operations research. Therefore, given the fact that fairness considerations have become of paramount importance in algorithm design, fairness in clustering has received significant attention from the research community. The literature on fair clustering has resulted in a collection of interesting fairness notions and elaborate algorithms. In this paper, we take a critical view of fair clustering, identifying a collection of ignored issues such as the lack of a clear utility characterization and the difficulty in accounting for the downstream effects of a fair clustering algorithm in machine learning settings. In some cases, we demonstrate examples where the application of a fair clustering algorithm can have significant negative impacts on social welfare. We end by identifying a collection of steps that would lead towards more impactful research in fair clustering.

Identifying misogyny using artificial intelligence is a form of combating online toxicity against women. However, the subjective nature of interpreting misogyny poses a significant challenge to model the phenomenon. In this paper, we propose a multitask learning approach that leverages the subjectivity of this task to enhance the performance of the misogyny identification systems. We incorporated diverse perspectives from annotators in our model design, considering gender and age across six profile groups, and conducted extensive experiments and error analysis using two language models to validate our four alternative designs of the multitask learning technique to identify misogynistic content in English tweets. The results demonstrate that incorporating various viewpoints enhances the language models' ability to interpret different forms of misogyny. This research advances content moderation and highlights the importance of embracing diverse perspectives to build effective online moderation systems.

While existing alignment paradigms have been integral in developing large language models (LLMs), LLMs often learn an averaged human preference and struggle to model diverse preferences across cultures, demographics, and communities. We propose Modular Pluralism, a modular framework based on multi-LLM collaboration for pluralistic alignment: it "plugs into" a base LLM a pool of smaller but specialized community LMs, where models collaborate in distinct modes to flexibility support three modes of pluralism: Overton, steerable, and distributional. Modular Pluralism is uniquely compatible with black-box LLMs and offers the modular control of adding new community LMs for previously underrepresented communities. We evaluate Modular Pluralism with six tasks and four datasets featuring questions/instructions with value-laden and perspective-informed responses. Extensive experiments demonstrate that Modular Pluralism advances the three pluralism objectives across six black-box and open-source LLMs. Further analysis reveals that LLMs are generally faithful to the inputs from smaller community LLMs, allowing seamless patching by adding a new community LM to better cover previously underrepresented communities.

Modern language models, while sophisticated, exhibit some inherent shortcomings, particularly in conversational settings. We claim that many of the observed shortcomings can be attributed to violation of one or more conversational principles. By drawing upon extensive research from both the social science and AI communities, we propose a set of maxims -- quantity, quality, relevance, manner, benevolence, and transparency -- for describing effective human-AI conversation. We first justify the applicability of the first four maxims (from Grice) in the context of human-AI interactions. We then argue that two new maxims, benevolence (concerning the generation of, and engagement with, harmful content) and transparency (concerning recognition of one's knowledge boundaries, operational constraints, and intents), are necessary for addressing behavior unique to modern human-AI interactions. We evaluate the degree to which various language models are able to understand these maxims and find that models possess an internal prioritization of principles that can significantly impact their ability to interpret the maxims accurately.