The growing volume of academic papers has made it increasingly difficult for researchers to efficiently extract key information. While large language models (LLMs) based agents are capable of automating question answering (QA) workflows for scientific papers, there still lacks a comprehensive and realistic benchmark to evaluate their capabilities. Moreover, training an interactive agent for this specific task is hindered by the shortage of high-quality interaction trajectories. In this work, we propose AirQA, a human-annotated comprehensive paper QA dataset in the field of artificial intelligence (AI), with 13,948 papers and 1,246 questions, that encompasses multi-task, multi-modal and instance-level evaluation. Furthermore, we propose ExTrActor, an automated framework for instruction data synthesis. With three LLM-based agents, ExTrActor can perform example generation and trajectory collection without human intervention. Evaluations of multiple open-source and proprietary models show that most models underperform on AirQA, demonstrating the quality of our dataset. Extensive experiments confirm that ExTrActor consistently improves the multi-turn tool-use capability of small models, enabling them to achieve performance comparable to larger ones.

--This paper focuses on the detection of potentially dangerous tendencies of social media users in an innovative multimodal way. We integrate Natural Language Processing (NLP) and Graph Neural Networks (GNNs) together. Firstly, we apply NLP on the user-generated text and conduct semantic analysis, sentiment recognition and keyword extraction to get subtle risk signals from social media posts. Meanwhile, we build a heterogeneous user relationship graph based on social interaction and propose a novel relational graph convolutional network to model user relationship, attention relationship and content dissemination path to discover some important structural information and user behaviors. Finally, we combine textual features extracted from these two models above with graph structural information, which provides a more robust and effective way to discover at-risk users. Our experiments on real social media datasets from different platforms show that our model can achieve significant improvement over single-modality methods.

Multi-channel audio alignment is a key requirement in bioacoustic monitoring, spatial audio systems, and acoustic localization. However, existing methods often struggle to address nonlinear clock drift and lack mechanisms for quantifying uncertainty. Traditional methods like Cross-correlation and Dynamic Time Warping assume simple drift patterns and provide no reliability measures. Meanwhile, recent deep learning models typically treat alignment as a binary classification task, overlooking inter-channel dependencies and uncertainty estimation. We introduce a method that combines cross-attention mechanisms with confidence-weighted scoring to improve multi-channel audio synchronization. We extend BEATs encoders with cross-attention layers to model temporal relationships between channels. We also develop a confidence-weighted scoring function that uses the full prediction distribution instead of binary thresholding. Our method achieved first place in the BioDCASE 2025 Task 1 challenge with 0.30 MSE average across test datasets, compared to 0.58 for the deep learning baseline. On individual datasets, we achieved 0.14 MSE on ARU data (77% reduction) and 0.45 MSE on zebra finch data (18% reduction). The framework supports probabilistic temporal alignment, moving beyond point estimates. While validated in a bioacoustic context, the approach is applicable to a broader range of multi-channel audio tasks where alignment confidence is critical. Code available on: https://github.com/Ragib-Amin-Nihal/BEATsCA

Similar to traditional SLAM systems, SLAM-F ormer comprises both a frontend and a backend that operate in tandem. The frontend processes sequential monocular images in real-time for incremental mapping and tracking, while the backend performs global refinement to ensure a geometrically consistent result. This alternating execution allows the frontend and backend to mutually promote one another, enhancing overall system performance. Comprehensive experimental results demonstrate that SLAM-F ormer achieves superior or highly competitive performance compared to state-of-the-art dense SLAM methods.

Survival analysis using whole-slide images (WSIs) is crucial in cancer research. Despite significant successes, pathology images typically only provide slide-level labels, which hinders the learning of discriminative representations from gigapixel WSIs. With the rapid advancement of high-throughput sequencing technologies, multimodal survival analysis integrating pathology images and genomics data has emerged as a promising approach. We propose a Multi-Expert Mamba (ME-Mamba) system that captures discriminative pathological and genomic features while enabling efficient integration of both modalities. This approach achieves complementary information fusion without losing critical information from individual modalities, thereby facilitating accurate cancer survival analysis. Specifically, we first introduce a Pathology Expert and a Genomics Expert to process unimodal data separately. Both experts are designed with Mamba architectures that incorporate conventional scanning and attention-based scanning mechanisms, allowing them to extract discriminative features from long instance sequences containing substantial redundant or irrelevant information. Second, we design a Synergistic Expert responsible for modality fusion. It explicitly learns token-level local correspondences between the two modalities via Optimal Transport, and implicitly enhances distribution consistency through a global cross-modal fusion loss based on Maximum Mean Discrepancy. The fused feature representations are then passed to a mamba backbone for further integration. Through the collaboration of the Pathology Expert, Genomics Expert, and Synergistic Expert, our method achieves stable and accurate survival analysis with relatively low computational complexity. Extensive experimental results on five datasets in The Cancer Genome Atlas (TCGA) demonstrate our state-of-the-art performance.

Most existing efforts on APA typically adopt segmental-level features as inputs and predict pronunciation scores at different granularities via hierarchical (or parallel) pronunciation modeling. This, however, inevitably causes assessments across linguistic levels (e.g., phone, word, and utterance) to rely solely on phoneme-level pronunciation features, nearly sidelining supra-segmental pronunciation cues. T o address this limitation, we introduce multi-task pre-training (MTP) for APA, a simple yet effective strategy that attempts to capture long-term temporal pronunciation cues while strengthening the intrinsic structures within an utterance via the objective of reconstructing input features. Specifically, for a phoneme-level encoder of an APA model, the proposed MTP strategy randomly masks segmental-level pronunciation features and reconstructs the masked ones based on their surrounding pronunciation context. Furthermore, current APA systems lack integration with automated speaking assessment (ASA), limiting holistic proficiency evaluation. Drawing on empirical studies and prior knowledge in ASA, our framework bridges this gap by incorporating handcrafted features (HCFs), such as fluency (speech rate, silence duration) and stress (pitch accent strength), derived from human-designed formulas via regressors to generate interpretable proficiency scores. Experiments on speechocean762 show improved pronunciation scoring and ASA proficiency correlation, enabling targeted training and comprehensive proficiency assessment. Index T erms--computer-assisted language learning, automatic pronunciation assessment, automated speaking assessment, multi-task learning.

Low Dynamic Range (LDR) to High Dynamic Range (HDR) image translation is a fundamental task in many computational vision problems. Numerous data-driven methods have been proposed to address this problem; however, they lack explicit modeling of illumination, lighting, and scene geometry in images. This limits the quality of the reconstructed HDR images. Since lighting and shadows interact differently with different materials, (e.g., specular surfaces such as glass and metal, and lambertian or diffuse surfaces such as wood and stone), modeling material-specific properties (e.g., specular and diffuse reflectance) has the potential to improve the quality of HDR image reconstruction. This paper presents PhysHDR, a simple yet powerful latent diffusion-based generative model for HDR image reconstruction. The denoising process is conditioned on lighting and depth information and guided by a novel loss to incorporate material properties of surfaces in the scene. The experimental results establish the efficacy of PhysHDR in comparison to a number of recent state-of-the-art methods.

We introduce seqBench, a parametrized benchmark for probing sequential reasoning limits in Large Language Models (LLMs) through precise, multi-dimensional control over several key complexity dimensions. seqBench allows systematic variation of (1) the logical depth, defined as the number of sequential actions required to solve the task; (2) the number of backtracking steps along the optimal path, quantifying how often the agent must revisit prior states to satisfy deferred preconditions (e.g., retrieving a key after encountering a locked door); and (3) the noise ratio, defined as the ratio between supporting and distracting facts about the environment. Our evaluations on state-of-the-art LLMs reveal a universal failure pattern: accuracy collapses exponentially beyond a model-specific logical depth. Unlike existing benchmarks, seqBench's fine-grained control facilitates targeted analyses of these reasoning failures, illuminating universal scaling laws and statistical limits, as detailed in this paper alongside its generation methodology and evaluation metrics. We find that even top-performing models systematically fail on seqBench's structured reasoning tasks despite minimal search complexity, underscoring key limitations in their commonsense reasoning capabilities. Designed for future evolution to keep pace with advancing models, the seqBench datasets are publicly released to spur deeper scientific inquiry into LLM reasoning, aiming to establish a clearer understanding of their true potential and current boundaries for robust real-world application.

Guardrails are critical for the safe deployment of Large Language Models (LLMs)-powered software. Unlike traditional rule-based systems with limited, predefined input-output spaces that inherently constrain unsafe behavior, LLMs enable open-ended, intelligent interactions--opening the door to jailbreak attacks through user inputs. Guardrails serve as a protective layer, filtering unsafe prompts before they reach the LLM. However, prior research shows that jailbreak attacks can still succeed over 70% of the time, even against advanced models like GPT-4o. While guardrails such as LlamaGuard report up to 95% accuracy, our preliminary analysis shows their performance can drop sharply--to as low as 12%--when confronted with unseen attacks. This highlights a growing software engineering challenge: how to build a post-deployment guardrail that adapts dynamically to emerging threats? To address this, we propose AdaptiveGuard, an adaptive guardrail that detects novel jailbreak attacks as out-of-distribution (OOD) inputs and learns to defend against them through a continual learning framework. Through empirical evaluation, AdaptiveGuard achieves 96% OOD detection accuracy, adapts to new attacks in just two update steps, and retains over 85% F1-score on in-distribution data post-adaptation, outperforming other baselines. These results demonstrate that AdaptiveGuard is a guardrail capable of evolving in response to emerging jailbreak strategies post deployment. We release our AdaptiveGuard and studied datasets at https://github.com/awsm-research/AdaptiveGuard to support further research.

Accurate assessment of intraventricular blood flow is essential for evaluating hemodynamic conditions in patients supported by Left Ventricular Assist Devices (LVADs). However, clinical imaging is either incompatible with LVADs or yields sparse, low-quality velocity data. While Computational Fluid Dynamics (CFD) simulations provide high-fidelity data, they are computationally intensive and impractical for routine clinical use. To address this, we propose LVADNet3D, a 3D convolutional autoencoder that reconstructs full-resolution intraventricular velocity fields from sparse velocity vector inputs. In contrast to a standard UNet3D model, LVADNet3D incorporates hybrid downsampling and a deeper encoder-decoder architecture with increased channel capacity to better capture spatial flow patterns. To train and evaluate the models, we generate a high-resolution synthetic dataset of intraventricular blood flow in LVAD-supported hearts using CFD simulations. We also investigate the effect of conditioning the models on anatomical and physiological priors. Across various input configurations, LVADNet3D outperforms the baseline UNet3D model, yielding lower reconstruction error and higher PSNR results.