The robotic intervention for individuals with Autism Spectrum Disorder (ASD) has generally used pre-defined scripts to deliver verbal content during one-to-one therapy sessions. This practice restricts the use of robots to limited, pre-mediated instructional curricula. In this paper, we increase robot autonomy in one such robotic intervention for children with ASD by implementing perspective-taking teaching. Our approach uses large language models (LLM) to generate verbal content as texts and then deliver it to the child via robotic speech. In the proposed pipeline, we teach perspective-taking through which our robot takes up three roles: initiator, prompter, and reinforcer. We adopted the GPT-2 + BART pipelines to generate social situations, ask questions (as initiator), and give options (as prompter) when required. The robot encourages the child by giving positive reinforcement for correct answers (as a reinforcer). In addition to our technical contribution, we conducted ten-minute sessions with domain experts simulating an actual perspective teaching session, with the researcher acting as a child participant. These sessions validated our robotic intervention pipeline through surveys, including those from NASA TLX and GodSpeed. We used BERTScore to compare our GPT-2 + BART pipeline with an all GPT-2 and found the performance of the former to be better. Based on the responses by the domain experts, the robot session demonstrated higher performance with no additional increase in mental or physical demand, temporal demand, effort, or frustration compared to a no-robot session. We also concluded that the domain experts perceived the robot as ideally safe, likable, and reliable.

Large language models (LLMs) integrated with retrieval-augmented generation (RAG) systems improve accuracy by leveraging external knowledge sources. However, recent research has revealed RAG's susceptibility to poisoning attacks, where the attacker injects poisoned texts into the knowledge database, leading to attacker-desired responses. Existing defenses, which predominantly focus on inference-time mitigation, have proven insufficient against sophisticated attacks. In this paper, we introduce RAGForensics, the first traceback system for RAG, designed to identify poisoned texts within the knowledge database that are responsible for the attacks. RAGForensics operates iteratively, first retrieving a subset of texts from the database and then utilizing a specially crafted prompt to guide an LLM in detecting potential poisoning texts. Empirical evaluations across multiple datasets demonstrate the effectiveness of RAGForensics against state-of-the-art poisoning attacks. This work pioneers the traceback of poisoned texts in RAG systems, providing a practical and promising defense mechanism to enhance their security. Our code is available at: this https URL

Retrieval-Augmented Generation (RAG) integrates external knowledge into large language models to improve response quality. However, recent work has shown that RAG systems are highly vulnerable to poisoning attacks, where malicious texts are inserted into the knowledge database to influence model outputs. While several defenses have been proposed, they are often circumvented by more adaptive or sophisticated attacks. This paper presents RAGOrigin, a black-box responsibility attribution framework designed to identify which texts in the knowledge database are responsible for misleading or incorrect generations. Our method constructs a focused attribution scope tailored to each misgeneration event and assigns a responsibility score to each candidate text by evaluating its retrieval ranking, semantic relevance, and influence on the generated response. The system then isolates poisoned texts using an unsupervised clustering method. We evaluate RAGOrigin across seven datasets and fifteen poisoning attacks, including newly developed adaptive poisoning strategies and multi-attacker scenarios. Our approach outperforms existing baselines in identifying poisoned content and remains robust under dynamic and noisy conditions. These results suggest that RAGOrigin provides a practical and effective solution for tracing the origins of corrupted knowledge in RAG systems. Our code is available at: this https URL

This report surveys the landscape of potential security threats from malicious uses of AI, and proposes ways to better forecast, prevent, and mitigate these threats. After analyzing the ways in which AI may influence the threat landscape in the digital, physical, and political domains, we make four high-level recommendations for AI researchers and other stakeholders. We also suggest several promising areas for further research that could expand the portfolio of defenses, or make attacks less effective or harder to execute. Finally, we discuss, but do not conclusively resolve, the long-term equilibrium of attackers and defenders.

3D plane reconstruction from a single image is a crucial yet challenging topic in 3D computer vision. Previous state-of-the-art (SOTA) methods have focused on training their system on a single dataset from either indoor or outdoor domain, limiting their generalizability across diverse testing data. In this work, we introduce a novel framework dubbed ZeroPlane, a Transformer-based model targeting zero-shot 3D plane detection and reconstruction from a single image, over diverse domains and environments. To enable data-driven models across multiple domains, we have curated a large-scale planar benchmark, comprising over 14 datasets and 560,000 high-resolution, dense planar annotations for diverse indoor and outdoor scenes. To address the challenge of achieving desirable planar geometry on multi-dataset training, we propose to disentangle the representation of plane normal and offset, and employ an exemplar-guided, classification-then-regression paradigm to learn plane and offset respectively. Additionally, we employ advanced backbones as image encoder, and present an effective pixel-geometry-enhanced plane embedding module to further facilitate planar reconstruction. Extensive experiments across multiple zero-shot evaluation datasets have demonstrated that our approach significantly outperforms previous methods on both reconstruction accuracy and generalizability, especially over in-the-wild data. Our code and data are available at: this https URL

We present a sample of 50 H-alpha detected broad-line active galactic nuclei (BLAGN) at redshifts 3.50), independent of the contributions of emission lines to the broadband photometry. We construct the black hole (BH) mass function at 3.5

Model merging is an emerging technique that integrates multiple models fine-tuned on different tasks to create a versatile model that excels in multiple domains. This scheme, in the meantime, may open up backdoor attack opportunities where one single malicious model can jeopardize the integrity of the merged model. Existing works try to demonstrate the risk of such attacks by assuming substantial computational resources, focusing on cases where the attacker can fully fine-tune the pre-trained model. Such an assumption, however, may not be feasible given the increasing size of machine learning models. In practice where resources are limited and the attacker can only employ techniques like Low-Rank Adaptation (LoRA) to produce the malicious model, it remains unclear whether the attack can still work and pose threats. In this work, we first identify that the attack efficacy is significantly diminished when using LoRA for fine-tuning. Then, we propose LoBAM, a method that yields high attack success rate with minimal training resources. The key idea of LoBAM is to amplify the malicious weights in an intelligent way that effectively enhances the attack efficacy. We demonstrate that our design can lead to improved attack success rate through extensive empirical experiments across various model merging scenarios. Moreover, we show that our method is highly stealthy and is difficult to detect and defend against.

JWST has revealed an abundance of low-luminosity active galactic nuclei (AGN) at high redshifts ($z > 3$), pushing the limits of black hole (BH) science in the early Universe. Results have claimed that these BHs are significantly more massive than expected from the BH mass-host galaxy stellar mass relation derived from the local Universe. We present a comprehensive census of the BH populations in the early Universe through a detailed stacking analysis of galaxy populations, binned by luminosity and redshift, using JWST spectroscopy from the CEERS, JADES, RUBIES, and GLASS extragalactic deep field surveys. Broad H$\alpha$ detections in $31\%$ of the stacked spectra (5/16 bins) imply median BH masses of $10^{5.21} - 10^{6.13}~ \rm{M_{\odot}}$ and the stacked SEDs of these bins indicate median stellar masses of $10^{7.84} - 10^{8.56} ~\rm{M_{\odot}}$. This suggests that the median galaxy hosts a BH that is at most a factor of 10 times over-massive compared to its host galaxy and lies closer to the locally derived $M_{BH}-M_*$ relation. We investigate the seeding properties of the inferred BHs and find that they can be well-explained by a light stellar remnant seed undergoing moderate Eddington accretion. Our results indicate that individual detections of AGN are more likely to sample the upper envelope of the $M_{BH}-M_*$ distribution, while stacking on ``normal" galaxies and searching for AGN signatures can overcome the selection bias of individual detections.

Human digital twins (HDTs) are dynamic, data-driven virtual representations of individuals, continuously updated with multimodal data to simulate, monitor, and predict health trajectories. By integrating clinical, physiological, behavioral, and environmental inputs, HDTs enable personalized diagnostics, treatment planning, and anomaly detection. This paper reviews current approaches to HDT modeling, with a focus on statistical and machine learning techniques, including recent advances in anomaly detection and failure prediction. It also discusses data integration, computational methods, and ethical, technological, and regulatory challenges in deploying HDTs for precision healthcare.

High resolution, ultraviolet imaging is often unavailable across the sky, even in heavily studied fields such as the Chandra Deep Field - South. The Habitable Worlds Observatory is one of two upcoming missions with the possibility of significant UV capabilities, and the only one early enough in development to consider suggestions to its design. In this paper, we conduct an initial study of how current common UV filter sets affect the results of spectral energy distribution fitting for the estimation of galaxy parameter. This initial look is intended to motivate the need for future, more robust, SED fitting of mock galaxies. We compare the broad near UV and far UV filters used by the GALEX mission to the three more narrow Swift UVOT filters. We find that the GALEX filters result in larger errors when calculating the UV beta parameter compared to UVOT, and provide little constraint on the star formation age of a galaxy. We further note the ability of the UVOT filters to investigate the 2175Å attenuation bump; GALEX has a reduced capacity to trace this same feature. Ultimately, we recommend that in order to optimize the effectiveness of HWO's ultraviolet capacity for transformative astrophysics, a minimum of a FUV filter with three medium band NUV filters should be adopted. This will combine the power of GALEX's wavelength range with the finer sampling of UVOT around an important dust feature.

We present the Cosmic Evolution Early Release Science Survey (CEERS) catalog, including space-based photometry, photometric redshifts, and physical parameters for more than 80,000 galaxies. The imaging used for this catalog comes from the CEERS survey, which has NIRCam coverage over ~100 sq. arcmin of the Extended Groth Strip (EGS) in seven filters from 1.15$\mu$m to 4.44$\mu$m. Alongside these data, we also include ancillary HST imaging in seven filters from 0.435$\mu$m to 1.6$\mu$m. We used Source Extractor with hot and cold detection settings to extract photometry. We derive photometric redshifts using the spectral energy distribution (SED) modeling code, LePHARE, and estimate their accuracy using spectroscopically confirmed galaxies out to $z\sim10$, with $\sigma_{NMAD}$ ranging from 0.035-0.073, depending strongly on galaxy magnitude and redshift. We compute stellar masses, star formation rates, and E(B-V) using three different SED fitting codes with different templates and assumptions about the galaxy star formation histories. All of these measurements, as well as the full mosaics in all filters, and redshift probability distribution functions, are made available via the CEERS DR1.0 data release.

Assistive technologies for people with visual impairments (PVI) have made significant advancements, particularly with the integration of artificial intelligence (AI) and real-time sensor technologies. However, current solutions often require PVI to switch between multiple apps and tools for tasks like image recognition, navigation, and obstacle detection, which can hinder a seamless and efficient user experience. In this paper, we present NaviGPT, a high-fidelity prototype that integrates LiDAR-based obstacle detection, vibration feedback, and large language model (LLM) responses to provide a comprehensive and real-time navigation aid for PVI. Unlike existing applications such as Be My AI and Seeing AI, NaviGPT combines image recognition and contextual navigation guidance into a single system, offering continuous feedback on the user's surroundings without the need for app-switching. Meanwhile, NaviGPT compensates for the response delays of LLM by using location and sensor data, aiming to provide practical and efficient navigation support for PVI in dynamic environments.

Open-set 3D object retrieval (3DOR) is an emerging task aiming to retrieve 3D objects of unseen categories beyond the training set. Existing methods typically utilize all modalities (i.e., voxels, point clouds, multi-view images) and train specific backbones before fusion. However, they still struggle to produce generalized representations due to insufficient 3D training data. Being contrastively pre-trained on web-scale image-text pairs, CLIP inherently produces generalized representations for a wide range of downstream tasks. Building upon it, we present a simple yet effective framework named Describe, Adapt and Combine (DAC) by taking only multi-view images for open-set 3DOR. DAC innovatively synergizes a CLIP model with a multi-modal large language model (MLLM) to learn generalized 3D representations, where the MLLM is used for dual purposes. First, it describes the seen category information to align with CLIP's training objective for adaptation during training. Second, it provides external hints about unknown objects complementary to visual cues during inference. To improve the synergy, we introduce an Additive-Bias Low-Rank adaptation (AB-LoRA), which alleviates overfitting and further enhances the generalization to unseen categories. With only multi-view images, DAC significantly surpasses prior arts by an average of +10.01\% mAP on four open-set 3DOR datasets. Moreover, its generalization is also validated on image-based and cross-dataset setups. Code is available at this https URL.

Millions of patients are already using large language model (LLM) chatbots for medical advice on a regular basis, raising patient safety concerns. This physician-led red-teaming study compares the safety of four publicly available chatbots--Claude by Anthropic, Gemini by Google, GPT-4o by OpenAI, and Llama3-70B by Meta--on a new dataset, HealthAdvice, using an evaluation framework that enables quantitative and qualitative analysis. In total, 888 chatbot responses are evaluated for 222 patient-posed advice-seeking medical questions on primary care topics spanning internal medicine, women's health, and pediatrics. We find statistically significant differences between chatbots. The rate of problematic responses varies from 21.6 percent (Claude) to 43.2 percent (Llama), with unsafe responses varying from 5 percent (Claude) to 13 percent (GPT-4o, Llama). Qualitative results reveal chatbot responses with the potential to lead to serious patient harm. This study suggests that millions of patients could be receiving unsafe medical advice from publicly available chatbots, and further work is needed to improve the clinical safety of these powerful tools.

We report the discovery of NGTS-11 c, a transiting warm Neptune ($P \approx 12.8$ d; $M_{p} = 1.2^{+0.3}_{-0.2} M_{\mathrm{Nep}}$; $R_{p} = 1.24 \pm 0.03 R_{\mathrm{Nep}}$), in an orbit interior to the previously reported transiting warm Saturn NGTS-11 b ($P \approx 35.5$ d). We also find evidence of a third outer companion orbiting the K-dwarf NGTS-11. We first detected transits of NGTS-11 c in TESS light curves and confirmed them with follow-up transits from NGTS and many other ground-based facilities. Radial-velocity monitoring with the HARPS and FEROS spectrographs revealed the mass of NGTS-11 c and provides evidence for a long-period companion ($P > 2300$ d; $M_{p} \sin i > 3.6 M_{\mathrm{Jup}}$). Taking into account the two additional bodies in our expanded datasets, we find that the mass of NGTS-11 b ($M_{p} = 0.63 \pm 0.09 M_{\mathrm{Sat}}$; $R_{p} = 0.97 \pm 0.02 R_{\mathrm{Sat}}$) is lower than previously reported ($M_{p} = 1.2 \pm 0.3 M_{\mathrm{Sat}}$). Given their near-circular and compact orbits, NGTS-11 c and b are unlikely to have reached their present locations via high-eccentricity migration. Instead, they probably either formed in situ or formed farther out and then underwent disk migration. A comparison of NGTS-11 with the eight other known systems hosting multiple well-characterized warm giants shows that it is most similar to Kepler-56. Finally, we find that the commonly used 10-day boundary between hot and warm Jupiters is empirically well supported.