Augmented Reality assistance are increasingly popular for supporting users with tasks like assembly and cooking. However, current practice typically provide reactive responses initialized from user requests, lacking consideration of rich contextual and user-specific information. To address this limitation, we propose a novel AR assistance system, Satori, that models both user states and environmental contexts to deliver proactive guidance. Our system combines the Belief-Desire-Intention (BDI) model with a state-of-the-art multi-modal large language model (LLM) to infer contextually appropriate guidance. The design is informed by two formative studies involving twelve experts. A sixteen within-subject study find that Satori achieves performance comparable to an designer-created Wizard-of-Oz (WoZ) system without relying on manual configurations or heuristics, thereby enhancing generalizability, reusability and opening up new possibilities for AR assistance.

Although large language models (LLMs) demonstrate impressive proficiency in various tasks, they present potential safety risks, such as `jailbreaks', where malicious inputs can coerce LLMs into generating harmful content bypassing safety alignments. In this paper, we delve into the ethical biases in LLMs and examine how those biases could be exploited for jailbreaks. Notably, these biases result in a jailbreaking success rate in GPT-4o models that differs by 20\% between non-binary and cisgender keywords and by 16\% between white and black keywords, even when the other parts of the prompts are identical. We introduce the concept of BiasJailbreak, highlighting the inherent risks posed by these safety-induced biases. BiasJailbreak generates biased keywords automatically by asking the target LLM itself, and utilizes the keywords to generate harmful output. Additionally, we propose an efficient defense method BiasDefense, which prevents jailbreak attempts by injecting defense prompts prior to generation. BiasDefense stands as an appealing alternative to Guard Models, such as Llama-Guard, that require additional inference cost after text generation. Our findings emphasize that ethical biases in LLMs can actually lead to generating unsafe output, and suggest a method to make the LLMs more secure and unbiased. To enable further research and improvements, we open-source our code and artifacts of BiasJailbreak, providing the community with tools to better understand and mitigate safety-induced biases in LLMs.

Despite significant advancements, Large Language Models (LLMs) exhibit blind spots that impair their ability to retrieve and process relevant contextual data effectively. We demonstrate that LLM performance in graph tasks with complexities beyond the "needle-in-a-haystack" scenario--where solving the problem requires cross-referencing and reasoning across multiple subproblems jointly--is influenced by the proximity of relevant information within the context, a phenomenon we term "lost-in-distance". We examine two fundamental graph tasks: identifying common connections between two nodes and assessing similarity among three nodes, and show that the model's performance in these tasks significantly depends on the relative positioning of common edges. We evaluate three publicly available LLMs using various graph encoding techniques that represent graph structures for LLM input. We propose a formulation for the lost-in-distance phenomenon and demonstrate that lost-in-distance and lost-in-the-middle phenomena occur independently. Results indicate that model accuracy can decline by up to 6x as the distance between node connections increases, independent of graph encoding and model size. Large Language Models (LLMs) have attained an unprecedented level of generality by leveraging scale and attention-based architectures (Kaplan et al., 2020; Vaswani, 2017). Additionally, LLMs are increasingly serving as essential and flexible building blocks for various user-facing machine learning and artificial intelligence applications beyond traditional language processing domains, such as recommendation systems (Geng et al., 2022), graph-related tasks (Wang et al., 2024), knowledge bases (AlKhamissi et al., 2022; Petroni et al., 2019), and more. These applications highlight the versatility of LLMs but also expose new challenges in handling domain-specific data encoded as textual input.

The exponential growth of data and advancements in big data technologies have created a demand for more efficient and automated approaches to data analysis and storytelling. However, automated data analysis systems still face challenges in leveraging large language models (LLMs) for data insight discovery, augmented analysis, and data storytelling. This paper introduces the Multidimensional Data Storytelling Framework (MDSF) based on large language models for automated insight generation and context-aware storytelling. The framework incorporates advanced preprocessing techniques, augmented analysis algorithms, and a unique scoring mechanism to identify and prioritize actionable insights. The use of fine-tuned LLMs enhances contextual understanding and generates narratives with minimal manual intervention. The architecture also includes an agent-based mechanism for real-time storytelling continuation control. Key findings reveal that MDSF outperforms existing methods across various datasets in terms of insight ranking accuracy, descriptive quality, and narrative coherence. The experimental evaluation demonstrates MDSF's ability to automate complex analytical tasks, reduce interpretive biases, and improve user satisfaction. User studies further underscore its practical utility in enhancing content structure, conclusion extraction, and richness of detail.

Large language models are increasingly being integrated into applications that shape the production and discovery of societal knowledge such as search, online education, and travel planning. As a result, language models will shape how people learn about, perceive and interact with global cultures making it important to consider whose knowledge systems and perspectives are represented in models. Recognizing this importance, increasingly work in Machine Learning and NLP has focused on evaluating gaps in global cultural representational distribution within outputs. However, more work is needed on developing benchmarks for cross-cultural impacts of language models that stem from a nuanced sociologically-aware conceptualization of cultural impact or harm. We join this line of work arguing for the need of metricizable evaluations of language technologies that interrogate and account for historical power inequities and differential impacts of representation on global cultures, particularly for cultures already under-represented in the digital corpora. We look at two concepts of erasure: omission: where cultures are not represented at all and simplification i.e. when cultural complexity is erased by presenting one-dimensional views of a rich culture. The former focuses on whether something is represented, and the latter on how it is represented. We focus our analysis on two task contexts with the potential to influence global cultural production. First, we probe representations that a language model produces about different places around the world when asked to describe these contexts. Second, we analyze the cultures represented in the travel recommendations produced by a set of language model applications. Our study shows ways in which the NLP community and application developers can begin to operationalize complex socio-cultural considerations into standard evaluations and benchmarks.

Transformer-based LLMs have achieved exceptional performance across a wide range of NLP tasks. However, the standard self-attention mechanism suffers from quadratic time complexity and linearly increased cache size. Sliding window attention (SWA) solves this problem by restricting the attention range to a fixed-size local context window. Nevertheless, SWA employs a uniform window size for each head in each layer, making it inefficient in capturing context of varying scales. To mitigate this limitation, we propose Multi-Scale Window Attention (MSWA) which applies diverse window sizes across heads and layers in the Transformer. It not only allows for different window sizes among heads within the same layer but also progressively increases window size allocation from shallow to deep layers, thus enabling the model to capture contextual information with different lengths and distances. Experimental results on language modeling and common-sense reasoning tasks substantiate that MSWA outperforms traditional local attention in both effectiveness and efficiency.

Transformers, driven by attention mechanisms, form the foundation of large language models (LLMs). As these models scale up, efficient GPU attention kernels become essential for high-throughput and low-latency inference. Diverse LLM applications demand flexible and high-performance attention solutions. We present FlashInfer: a customizable and efficient attention engine for LLM serving. FlashInfer tackles KV-cache storage heterogeneity using block-sparse format and composable formats to optimize memory access and reduce redundancy. It also offers a customizable attention template, enabling adaptation to various settings through Just-In-Time (JIT) compilation. Additionally, FlashInfer's load-balanced scheduling algorithm adjusts to dynamism of user requests while maintaining compatibility with CUDAGraph which requires static configuration. FlashInfer have been integrated into leading LLM serving frameworks like SGLang, vLLM and MLC-Engine. Comprehensive kernel-level and end-to-end evaluations demonstrate FlashInfer's ability to significantly boost kernel performance across diverse inference scenarios: compared to state-of-the-art LLM serving solutions, FlashInfer achieve 29-69% inter-token-latency reduction compared to compiler backends for LLM serving benchmark, 28-30% latency reduction for long-context inference, and 13-17% speedup for LLM serving with parallel generation.

In psychological practice, standardized questionnaires serve as essential tools for assessing mental constructs (e.g., attitudes, traits, and emotions) through structured questions (aka items). With the increasing prevalence of social media platforms where users share personal experiences and emotions, researchers are exploring computational methods to leverage this data for rapid mental health screening. In this study, we propose a novel adaptive Retrieval-Augmented Generation (RAG) approach that completes psychological questionnaires by analyzing social media posts. Our method retrieves the most relevant user posts for each question in a psychological survey and uses Large Language Models (LLMs) to predict questionnaire scores in a zero-shot setting. Our findings are twofold. First we demonstrate that this approach can effectively predict users' responses to psychological questionnaires, such as the Beck Depression Inventory II (BDI-II), achieving performance comparable to or surpassing state-of-the-art models on Reddit-based benchmark datasets without relying on training data. Second, we show how this methodology can be generalized as a scalable screening tool, as the final assessment is systematically derived by completing standardized questionnaires and tracking how individual item responses contribute to the diagnosis, aligning with established psychometric practices.

Images generated by text-to-image (T2I) models often exhibit visual biases and stereotypes of concepts such as culture and profession. Existing quantitative measures of stereotypes are based on statistical parity that does not align with the sociological definition of stereotypes and, therefore, incorrectly categorizes biases as stereotypes. Instead of oversimplifying stereotypes as biases, we propose a quantitative measure of stereotypes that aligns with its sociological definition. We then propose OASIS to measure the stereotypes in a generated dataset and understand their origins within the T2I model. S to measure spectral variance in the images along a stereotypical attribute. OASIS also includes two methods to understand the origins of stereotypes in T2I models: (U1) StOP to discover attributes that the T2I model internally associates with a given concept, and (U2) SPI to quantify the emergence of stereotypical attributes in the latent space of the T2I model during image generation. Despite the considerable progress in image fidelity, using OASIS, we conclude that newer T2I models such as FLUX.1 and SDv3 contain strong stereotypical predispositions about concepts and still generate images with widespread stereotypical attributes. S measures the variance of images along these attributes. In a sociological context, stereotypes are generalized beliefs or assumptions about a particular group of people, things, or categories (Bordalo et al., 2016). For instance, consider the images in Figure 1 generated by FLUX.1 (BlackForestLabs, 2024), SDv3 (Esser et al., 2024), and SDv2 (Rombach et al., 2022) for the prompt "A photo of a/an person". There are clear portrayals of ethnic stereotypes in attributes such as clothing, skin tone, and facial features across different nationalities, despite no references to such attributes in the prompt. For example, the model consistently depicts an Iranian person as a middle-aged or senior with a long beard, wearing a turban, and dressed in religious attire, reinforcing harmful stereotypical representations about people with Iranian nationality. Besides being demographically incorrect, stereotypical biases in these models can lead to broader harm.

Designing structured visuals such as presentation slides is essential for communicative needs, necessitating both content creation and visual planning skills. In this work, we tackle the challenge of automated slide generation, where models produce slide presentations from natural language (NL) instructions. We first introduce the SlidesBench benchmark, the first benchmark for slide generation with 7k training and 585 testing examples derived from 310 slide decks across 10 domains. SlidesBench supports evaluations that are (i)reference-based to measure similarity to a target slide, and (ii)reference-free to measure the design quality of generated slides alone. We benchmark end-to-end image generation and program generation methods with a variety of models, and find that programmatic methods produce higher-quality slides in user-interactable formats. Built on the success of program generation, we create AutoPresent, an 8B Llama-based model trained on 7k pairs of instructions paired with code for slide generation, and achieve results comparable to the closed-source model GPT-4o. We further explore iterative design refinement where the model is tasked to self-refine its own output, and we found that this process improves the slide's quality. We hope that our work will provide a basis for future work on generating structured visuals.