Variational Dual-path Attention Network for CSI-Based Gesture Recognition

Wi-Fi gesture recognition based on Channel State Information (CSI) is challenged by high-dimensional noise and resource constraints on edge devices. Prevailing end-to-end models tightly couple feature extraction with classification, overlooking the inherent time-frequency sparsity of CSI and leading to redundancy and poor generalization. To address this, this paper proposes a lightweight feature preprocessing module--the Variational Dual-path Attention Network (VDAN). It performs structured feature refinement through frequency-domain filtering and temporal detection. Variational inference is introduced to model the uncertainty in attention weights, thereby enhancing robustness to noise. The design principles of the module are explained from the perspectives of the information bottleneck and regularization. Experiments on a public dataset demonstrate that the learned attention weights align with the physical sparse characteristics of CSI, verifying its interpretability. This work provides an efficient and explainable front-end processing solution for resource-constrained wireless sensing systems.

UniRoute: Unified Routing Mixture-of-Experts for Modality-Adaptive Remote Sensing Change Detection

Current remote sensing change detection (CD) methods mainly rely on specialized models, which limits the scalability toward modality-adaptive Earth observation. For homogeneous CD, precise boundary delineation relies on fine-grained spatial cues and local pixel interactions, whereas heterogeneous CD instead requires broader contextual information to suppress speckle noise and geometric distortions. Moreover, difference operator (e.g., subtraction) works well for aligned homogeneous images but introduces artifacts in cross-modal or geometrically misaligned scenarios. Across different modality settings, specialized models based on static backbones or fixed difference operations often prove insufficient. To address this challenge, we propose UniRoute, a unified framework for modality-adaptive learning by reformulating feature extraction and fusion as conditional routing problems. We introduce an Adaptive Receptive Field Routing MoE (AR2-MoE) module to disentangle local spatial details from global semantic context, and a Modality-Aware Difference Routing MoE (MDR-MoE) module to adaptively select the most suitable fusion primitive at each pixel. In addition, we propose a Consistency-Aware Self-Distillation (CASD) strategy that stabilizes unified training under data-scarce heterogeneous settings by enforcing multi-level consistency. Extensive experiments on five public datasets demonstrate that UniRoute achieves strong overall performance, with a favorable accuracy-efficiency trade-off under a unified deployment setting.

Ontology-aligned structuring and reuse of multimodal materials data and workflows towards automatic reproduction

Reproducibility of computational results remains a challenge in materials science, as simulation workflows and parameters are often reported only in unstructured text and tables. While literature data are valuable for validation and reuse, the lack of machine-readable workflow descriptions prevents large-scale curation and systematic comparison. Existing text-mining approaches are insufficient to extract complete computational workflows with their associated parameters. An ontology-driven, large language model (LLM)-assisted framework is introduced for the automated extraction and structuring of computational workflows from the literature. The approach focuses on density functional theory-based stacking fault energy (SFE) calculations in hexagonal close-packed magnesium and its binary alloys, and uses a multi-stage filtering strategy together with prompt-engineered LLM extraction applied to method sections and tables. Extracted information is unified into a canonical schema and aligned with established materials ontologies (CMSO, ASMO, and PLDO), enabling the construction of a knowledge graph using atomRDF. The resulting knowledge graph enables systematic comparison of reported SFE values and supports the structured reuse of computational protocols. While full computational reproducibility is still constrained by missing or implicit metadata, the framework provides a foundation for organizing and contextualizing published results in a semantically interoperable form, thereby improving transparency and reusability of computational materials data.

PaperGuide: Making Small Language-Model Paper-Reading Agents More Efficient

The accelerating growth of the scientific literature makes it increasingly difficult for researchers to track new advances through manual reading alone. Recent progress in large language models (LLMs) has therefore spurred interest in autonomous agents that can read scientific papers and extract task-relevant information. However, most existing approaches rely either on heavily engineered prompting or on a conventional SFT-RL training pipeline, both of which often lead to excessive and low-yield exploration. Drawing inspiration from cognitive science, we propose PaperCompass, a framework that mitigates these issues by separating high-level planning from fine-grained execution. PaperCompass first drafts an explicit plan that outlines the intended sequence of actions, and then performs detailed reasoning to instantiate each step by selecting the parameters for the corresponding function calls. To train such behavior, we introduce Draft-and-Follow Policy Optimization (DFPO), a tailored RL method that jointly optimizes both the draft plan and the final solution. DFPO can be viewed as a lightweight form of hierarchical reinforcement learning, aimed at narrowing the `knowing-doing' gap in LLMs. We provide a theoretical analysis that establishes DFPO's favorable optimization properties, supporting a stable and reliable training process. Experiments on paper-based question answering (Paper-QA) benchmarks show that PaperCompass improves efficiency over strong baselines without sacrificing performance, achieving results comparable to much larger models.

Does medical specialization of VLMs enhance discriminative power?: A comprehensive investigation through feature distribution analysis

This study investigates the feature representations produced by publicly available open source medical vision-language models (VLMs). While medical VLMs are expected to capture diagnostically relevant features, their learned representations remain underexplored, and standard evaluations like classification accuracy do not fully reveal if they acquire truly discriminative, lesion-specific features. Understanding these representations is crucial for revealing medical image structures and improving downstream tasks in medical image analysis. This study aims to investigate the feature distributions learned by medical VLMs and evaluate the impact of medical specialization. We analyze the feature distribution of multiple image modalities extracted by some representative medical VLMs across lesion classification datasets on multiple modalities. These distributions were compared them with non-medical VLMs to assess the domain-specific medical training. Our experiments showed that medical VLMs can extract discriminative features that are effective for medical classification tasks. Moreover, it was found that non-medical VLMs with recent improvement with contextual enrichment such as LLM2CLIP produce more refined feature representations. Our results imply that enhancing text encoder is more crucial than training intensively on medical images when developing medical VLMs. Notably, non-medical models are particularly vulnerable to biases introduced by overlaied text strings on images. These findings underscore the need for careful consideration on model selection according to downstream tasks besides potential risks in inference due to background biases such as textual information in images.

Incorporating Q&A Nuggets into Retrieval-Augmented Generation

RAGE systems integrate ideas from automatic evaluation (E) into Retrieval-augmented Generation (RAG). As one such example, we present Crucible, a Nugget-Augmented Generation System that preserves explicit citation provenance by constructing a bank of Q&A nuggets from retrieved documents and uses them to guide extraction, selection, and report generation. Reasoning on nuggets avoids repeated information through clear and interpretable Q&A semantics - instead of opaque cluster abstractions - while maintaining citation provenance throughout the entire generation process. Evaluated on the TREC NeuCLIR 2024 collection, our Crucible system substantially outperforms Ginger, a recent nugget-based RAG system, in nugget recall, density, and citation grounding.

Fusion Segment Transformer: Bi-Directional Attention Guided Fusion Network for AI-Generated Music Detection

With the rise of generative AI technology, anyone can now easily create and deploy AI-generated music, which has heightened the need for technical solutions to address copyright and ownership issues. While existing works mainly focused on short-audio, the challenge of full-audio detection, which requires modeling long-term structure and context, remains insufficiently explored. To address this, we propose an improved version of the Segment Transformer, termed the Fusion Segment Transformer. As in our previous work, we extract content embeddings from short music segments using diverse feature extractors. Furthermore, we enhance the architecture for full-audio AI-generated music detection by introducing a Gated Fusion Layer that effectively integrates content and structural information, enabling the capture of long-term context. Experiments on the SONICS and AIME datasets show that our approach outperforms the previous model and recent baselines, achieving state-of-the-art results in AI-generated music detection.

Balancing Classification and Calibration Performance in Decision-Making LLMs via Calibration Aware Reinforcement Learning

Large language models (LLMs) are increasingly deployed in decision-making tasks, where not only accuracy but also reliable confidence estimates are essential. Well-calibrated confidence enables downstream systems to decide when to trust a model and when to defer to fallback mechanisms. In this work, we conduct a systematic study of calibration in two widely used fine-tuning paradigms: supervised fine-tuning (SFT) and reinforcement learning with verifiable rewards (RLVR). We show that while RLVR improves task performance, it produces extremely overconfident models, whereas SFT yields substantially better calibration, even under distribution shift, though with smaller performance gains. Through targeted experiments, we diagnose RLVR's failure, showing that decision tokens act as extraction steps of the decision in reasoning traces and do not carry confidence information, which prevents reinforcement learning from surfacing calibrated alternatives. Based on this insight, we propose a calibration-aware reinforcement learning formulation that directly adjusts decision-token probabilities. Our method preserves RLVR's accuracy level while mitigating overconfidence, reducing ECE scores up to 9 points.

From Tags to Trees: Structuring Fine-Grained Knowledge for Controllable Data Selection in LLM Instruction Tuning

Effective and controllable data selection is critical for LLM instruction tuning, especially with massive open-source datasets. Existing approaches primarily rely on instance-level quality scores, or diversity metrics based on embedding clusters or semantic tags. However, constrained by the flatness of embedding spaces or the coarseness of tags, these approaches overlook fine-grained knowledge and its intrinsic hierarchical dependencies, consequently hindering precise data valuation and knowledge-aligned sampling. To address this challenge, we propose Tree-aware Aligned Global Sampling (TAGS), a unified framework that leverages a knowledge tree built from fine-grained tags, thereby enabling joint control of global quality, diversity, and target alignment. Using an LLM-based tagger, we extract atomic knowledge concepts, which are organized into a global tree through bottom-up hierarchical clustering. By grounding data instances onto this tree, a tree-aware metric then quantifies data quality and diversity, facilitating effective sampling. Our controllable sampling strategy maximizes tree-level information gain and enforces leaf-level alignment via KL-divergence for specific domains. Extensive experiments demonstrate that TAGS significantly outperforms state-of-the-art baselines. Notably, it surpasses the full-dataset model by \textbf{+5.84\%} using only \textbf{5\%} of the data, while our aligned sampling strategy further boosts average performance by \textbf{+4.24\%}.

Event Classification by Physics-informed Inpainting for Distributed Multichannel Acoustic Sensor with Partially Degraded Channels

Distributed multichannel acoustic sensing (DMAS) enables large-scale sound event classification (SEC), but performance drops when many channels are degraded and when sensor layouts at test time differ from training layouts. We propose a learning-free, physics-informed inpainting frontend based on reverse time migration (RTM). In this approach, observed multichannel spectrograms are first back-propagated on a 3D grid using an analytic Green's function to form a scene-consistent image, and then forward-projected to reconstruct inpainted signals before log-mel feature extraction and Transformer-based classification. We evaluate the method on ESC-50 with 50 sensors and three layouts (circular, linear, right-angle), where per-channel SNRs are sampled from -30 to 0 dB. Compared with an AST baseline, scaling-sparsemax channel selection, and channel-swap augmentation, the proposed RTM frontend achieves the best or competitive accuracy across all layouts, improving accuracy by 13.1 points on the right-angle layout (from 9.7% to 22.8%). Correlation analyses show that spatial weights align more strongly with SNR than with channel--source distance, and that higher SNR--weight correlation corresponds to higher SEC accuracy. These results demonstrate that a reconstruct-then-project, physics-based preprocessing effectively complements learning-only methods for DMAS under layout-open configurations and severe channel degradation.