Joaquin Oliver was 17 years old when he was shot in the hallway of his high school. An older teenager, expelled some months previously, had opened fire with a high-powered rifle on Valentine's Day in what became America's deadliest high school shooting. Seven years on, Joaquin says he thinks it's important to talk about what happened on that day in Parkland, Florida, "so that we can create a safer future for everyone". But sadly, what happened to Joaquin that day is that he died. The oddly metallic voice speaking to the ex-CNN journalist Jim Acosta in an interview on Substack this week was actually that of a digital ghost: an AI, trained on the teenager's old social media posts at the request of his parents, who are using it to bolster their campaign for tougher gun controls.

Large Language Models (LLMs) have started to demonstrate the ability to persuade humans, yet our understanding of how this dynamic transpires is limited. Recent work has used linear probes, lightweight tools for analyzing model representations, to study various LLM skills such as the ability to model user sentiment and political perspective. Motivated by this, we apply probes to study persuasion dynamics in natural, multi-turn conversations. We leverage insights from cognitive science to train probes on distinct aspects of persuasion: persuasion success, persuadee personality, and persuasion strategy. Despite their simplicity, we show that they capture various aspects of persuasion at both the sample and dataset levels. For instance, probes can identify the point in a conversation where the persuadee was persuaded or where persuasive success generally occurs across the entire dataset. We also show that in addition to being faster than expensive prompting-based approaches, probes can do just as well and even outperform prompting in some settings, such as when uncovering persuasion strategy. This suggests probes as a plausible avenue for studying other complex behaviours such as deception and manipulation, especially in multi-turn settings and large-scale dataset analysis where prompting-based methods would be computationally inefficient.

This paper introduces a comprehensive benchmark for evaluating how Large Language Models (LLMs) respond to linguistic shibboleths: subtle linguistic markers that can inadvertently reveal demographic attributes such as gender, social class, or regional background. Through carefully constructed interview simulations using 100 validated question-response pairs, we demonstrate how LLMs systematically penalize certain linguistic patterns, particularly hedging language, despite equivalent content quality. Our benchmark generates controlled linguistic variations that isolate specific phenomena while maintaining semantic equivalence, which enables the precise measurement of demographic bias in automated evaluation systems. We validate our approach along multiple linguistic dimensions, showing that hedged responses receive 25.6% lower ratings on average, and demonstrate the benchmark's effectiveness in identifying model-specific biases. This work establishes a foundational framework for detecting and measuring linguistic discrimination in AI systems, with broad applications to fairness in automated decision-making contexts.

--Spectral clustering is a leading clustering method. Two of its major shortcomings are the disjoint optimization process and the limited representation capacity. T o address these issues, we propose a deep spectral clustering model (named BootSC), which jointly learns all stages of spectral clustering-- affinity matrix construction, spectral embedding, and k -means clustering--using a single network in an end-to-end manner . Moreover, a semantically-consistent orthogonal re-parameterization technique is introduced to or-thogonalize spectral embeddings, significantly enhancing the discrimination capability. Experimental results indicate that BootSC achieves state-of-the-art clustering performance. For example, it accomplishes a notable 16% NMI improvement over the runner-up method on the challenging ImageNet-Dogs dataset. EEP clustering models aim to detect underlying cluster structures within unlabelled data. To train these models, creating effective and efficient supervision signals is necessary. Inadequate supervision could result in excessive computational costs [1], training instability [2], and degenerate results [3]. Classical deep clustering models [5], [6], [7], [8], [9], [10] commonly adopt cluster assignments obtained by k -means on data representations as training supervision. A major challenge with this k -means-style supervision is that data representations are assumed to follow simple isotropic Gaussian distributions.

Reinforcement learning-based retrieval-augmented generation (RAG) methods enhance the reasoning abilities of large language models (LLMs). However, most rely only on final-answer rewards, overlooking intermediate reasoning quality. This paper analyzes existing RAG reasoning models and identifies three main failure patterns: (1) information insufficiency, meaning the model fails to retrieve adequate support; (2) faulty reasoning, where logical or content-level flaws appear despite sufficient information; and (3) answer-reasoning inconsistency, where a valid reasoning chain leads to a mismatched final answer. We propose TIRESRAG-R1, a novel framework using a think-retrieve-reflect process and a multi-dimensional reward system to improve reasoning and stability. TIRESRAG-R1 introduces: (1) a sufficiency reward to encourage thorough retrieval; (2) a reasoning quality reward to assess the rationality and accuracy of the reasoning chain; and (3) a reflection reward to detect and revise errors. It also employs a difficulty-aware reweighting strategy and training sample filtering to boost performance on complex tasks. Experiments on four multi-hop QA datasets show that TIRESRAG-R1 outperforms prior RAG methods and generalizes well to single-hop tasks. The code and data are available at: https://github.com/probe2/TIRESRAG-R1.

As AI systems become integral to knowledge - intensive work, questions arise not only about their functionality but also their epistemic roles in human - AI interaction. While HCI research has proposed various AI role typologies, it often overlooks how AI resh apes users' roles as knowledge contributors. This study examines how users form epistemic relationships with AI -- how they assess, trust, and collaborate with it in research and teaching contexts. Based on 31 interviews with academics across disciplines, we developed a five - part codebook and identified five relationship types: Instrumental Reliance, Contingent Delegation, Co - agency Collaboration, Authority Displacement, and Epistemic Abstention. These reflect variations in trust, assessment modes, tasks, and human epistemic status. Our findings show that epistemic roles are dynamic and context dependent . We argue for shifting beyond static metaphors of AI toward a more nuanced framework that captures how humans and AI co - construct knowledge, enriching HCI's understanding of the relational and normative dimensions of AI use.

Recent advancements in large multimodal models have provided blind or visually impaired (BVI) individuals with new capabilities to interpret and engage with the real world through interactive systems that utilize live video feeds. However, the potential benefits and challenges of such capabilities to support diverse real-world assistive tasks remain unclear. In this paper, we present findings from an exploratory study with eight BVI participants. Participants used ChatGPT's Advanced Voice with Video, a state-of-the-art live video AI released in late 2024, in various real-world scenarios, from locating objects to recognizing visual landmarks, across unfamiliar indoor and outdoor environments. Our findings indicate that current live video AI effectively provides guidance and answers for static visual scenes but falls short in delivering essential live descriptions required in dynamic situations. Despite inaccuracies in spatial and distance information, participants leveraged the provided visual information to supplement their mobility strategies. Although the system was perceived as human-like due to high-quality voice interactions, assumptions about users' visual abilities, hallucinations, generic responses, and a tendency towards sycophancy led to confusion, distrust, and potential risks for BVI users. Based on the results, we discuss implications for assistive video AI agents, including incorporating additional sensing capabilities for real-world use, determining appropriate intervention timing beyond turn-taking interactions, and addressing ecological and safety concerns.

Emotional and mental well-being are vital components of quality of life, and with the rise of smart devices like smartphones, wearables, and artificial intelligence (AI), new opportunities for monitoring emotions in everyday settings have emerged. However, for AI algorithms to be effective, they require high-quality data and accurate annotations. As the focus shifts towards collecting emotion data in real-world environments to capture more authentic emotional experiences, the process of gathering emotion annotations has become increasingly complex. This work explores the challenges of everyday emotion data collection from the perspectives of key stakeholders. We collected 75 survey responses, performed 32 interviews with the public, and 3 focus group discussions (FGDs) with 12 mental health professionals. The insights gained from a total of 119 stakeholders informed the development of our framework, AnnoSense, designed to support everyday emotion data collection for AI. This framework was then evaluated by 25 emotion AI experts for its clarity, usefulness, and adaptability. Lastly, we discuss the potential next steps and implications of AnnoSense for future research in emotion AI, highlighting its potential to enhance the collection and analysis of emotion data in real-world contexts.

As Will points out, there were two recent wins for OpenAI in its efforts to build AI that outcompetes humans. Its models took second place at a top-level coding competition and--alongside those from Google DeepMind--achieved gold-medal-level results in the 2025 International Math Olympiad. People who believe that AI doesn't pose genuine competition to human-level intelligence might actually take some comfort in that. AI is good at the mathematical and analytical, which are on full display in olympiads and coding competitions. That doesn't mean it's any good at grappling with the messiness of human emotions, making hard decisions, or creating art that resonates with anyone. But that distinction--between machine-like reasoning and the ability to think creatively--is not one OpenAI's heads of research are inclined to make.

High-performance supercomputing--once the exclusive domain of scientific research--is now a strategic resource for training increasingly complex artificial intelligence models. This convergence of AI and HPC is redefining not only these technologies, but also the ways in which knowledge is produced, and takes a strategic position in the global landscape. To discuss how HPC is evolving, in July WIRED caught up with Jack Dongarra, a US computer scientist who has been a key contributor to the development of HPC software over the past four decades--so much so that in 2021 he earned the prestigious Turing Award. The meeting took place at the 74th Nobel Laureate Meeting in Lindau, Germany, which brought together dozens of Nobel laureates as well as more than 600 emerging scientists from around the world. This interview has been edited for length and clarity.