Buildings are the largest engineered mass on earth. Cement-based supercapacitors promise distributed energy storage in structural elements, yet their design is governed by a coupled ionic-electronic transport problem that remains poorly resolved: conductive fillers that maximize electrochemical performance simultaneously dilute the load-bearing matrix. Here, these competing contributions are disentangled in carbon–cement composites, and the resulting framework is used to define the multifunctional optimum. Employing cassava (Manihot esculenta) leaf-derived biochar, an agricultural residue, as a model intrinsic-porosity carbon, the total conductivity is resolved into a finite ionic background and an electronic percolation branch, with an apparent threshold pc≈13.3 wt% and a critical exponent t≈2 consistent with a three-dimensional random conductive network. Micro-computed tomography, electrochemical impedance spectroscopy, and a Bruggeman-type ionic-tortuosity estimate are combined into a single semi-analytical model that links microstructure, transport, and mechanical response. The composites deliver areal capacitances approaching 180mFcm−2, and power densities of 3–4kWkg−1 on an active-carbon basis. On a total-device basis, capacitance reaches ≈ 35 mF g−1 and ≈71 mF cm−3 (BC80). Across loading, electronic connectivity, ionic accessibility, and paste-level strength (8–12 MPa at BC20-BC40) trade off in the manner the model predicts, identifying BC40–BC60 as the multifunctional window. The central result is a transferable design framework providing quantitative design guidance for the present material system for balancing percolation, ionic transport, and matrix continuity in structural energy-storage cement.
Repetitive head impacts pose a significant threat to brain health. However, their injury accumulation mechanism remains elusive. Here, we study how traumatic axonal injury might accumulate based on typical impact severities and frequencies obtained from University Men’s Ice Hockey. From N = 994 impact simulations over a season (N = 22 athletes), we identified 50th (‘‘mild’’, 7.0
We investigate coupled fluid flow, heat transfer, species transport, and surface reaction in a single reactive pore embedded within a low-permeability matrix layer separating adjacent aquifer strata, motivated by aquifer thermal energy storage (ATES) applications. Building on fracture-scale reactive-transport simulations, we develop a pore-scale model that resolves reaction-driven pore evolution, and assess whether such evolution can compromise the impermeability assumption employed in homogenized ATES models. Analysis of a spatially independent reduction reveals a robust thermochemical structure in which the dominant reaction mode is governed by thermal forcing, while reaction and matrix-fluid heat exchange control transient relaxation toward equilibrium. Numerical simulations of the full pore-scale system of partial differential equations (PDE) show how axial transport redistributes these local dynamics without altering their underlying structure. Using time-dependent thermal and chemical forcing extracted from a fracture-scale model, we identify parameter regimes in which reaction-driven pore evolution remains insufficient to exceed experimentally reported permeability thresholds. These results delineate a regime of validity for the impermeability assumption in ATES modeling and provide a mechanistic link between fracture-scale forcing and pore-scale transport processes.
Monaural multi-speaker automatic speech recognition (ASR) remains challenging due to data scarcity and the intrinsic difficulty of recognizing and attributing words to individual speakers, particularly in overlapping speech. Recent advances have driven the shift from cascade systems to end-to-end (E2E) architectures, which reduce error propagation and better exploit the synergy between speech content and speaker identity. Despite rapid progress in E2E multi-speaker ASR, the field lacks a comprehensive review of recent developments. This survey provides a systematic taxonomy of E2E neural approaches for multi-speaker ASR, highlighting recent advances and comparative analysis. Specifically, we analyze: (1) architectural paradigms (single-input-multiple-output (SIMO) vs. single-input-single-output (SISO)) for pre-segmented audio, analyzing their distinct characteristics and trade-offs; (2) recent architectural and algorithmic improvements based on these two paradigms, including multi-modal inputs; (3) extensions to long-form speech, including segmentation strategy and speaker-consistent hypothesis stitching. Further, we (4) evaluate and compare methods across standard benchmarks. We conclude with a discussion of open challenges and future research directions towards building robust and scalable multi-speaker ASR.
Accurate recall from large-scale memories remains a core challenge for memory-augmented AI assistants performing question answering (QA), especially in similarity-dense scenarios where existing methods mainly rely on semantic distance to the query for retrieval. Inspired by how humans link information associatively, we propose AssoMem, a novel framework constructing an associative memory graph that anchors dialogue utterances to automatically extracted clues. This structure provides a rich organizational view of the conversational context and facilitates importance-aware ranking. Further, AssoMem integrates multi-dimensional retrieval signals—relevance, importance, and temporal alignment—using an adaptive mutual information (MI)-driven fusion strategy. Extensive experiments across three benchmarks and a newly introduced dataset, MeetingQA, demonstrate that AssoMem consistently outperforms state-of-the-art baselines, verifying its superiority in context-aware memory recall.
