遼寧科技大學

Multimodal learning aims to exploit cross-modal complementarity by integrating heterogeneous modalities such as text, audio, and vision. However, multimodal models often suffer from the imbalanced multimodal learning problem, where the dominant modality monopolizes optimization resources while weaker modalities remain insufficiently learned. Empirical studies show that, unlike unimodal training, gradients across modalities exhibit highly similar evolution trends in multimodal settings, even with existing balancing strategies, indicating that gradient magnitude adjustment alone fails to address the underlying optimization dynamics. In this work, we reanalyze imbalanced multimodal learning from an optimization perspective by incorporating Cognitive Load Theory. We show that, under limited model capacity, jointly optimizing multiple modalities induces a persistent high load state, where gradients from different modalities compete within a constrained update subspace, resulting in gradient conflicts and a persistent imbalanced state. Inspired by human cognitive behavior, we propose a multimodal balanced learning framework that integrates a plug-and-play multimodal alternating learning strategy to alleviate the learning load of simultaneous multimodal optimization and a Gaussian noise based optimizer to prevent overly similar gradient magnitude change trends. Extensive experiments on IEMOCAP, CMU-MOSEI, and AVE demonstrate the rationality and effectiveness of the proposed framework.

This study utilized discarded bovine bones as raw material to prepare K2CO3-modified bone char catalyst (BC-K) via an impregnation-pyrolysis process. The catalytic pyrolysis gasification performance of this catalyst was systematically evaluated for three types of carbon-hydrogen solid waste: cotton stalks, polypropylene plastic, and corn starch. XRD and density functional theory (DFT) analyses revealed successful co-doping of K+ and CO32- ions into the hydroxyapatite lattice, forming a new crystalline phase Ca9K(PO4)5(CO3)(OH)2. The lattice parameters expanded from a = b = 9.43 & Aring;, c = 6.88 & Aring; to a = b = 9.45 & Aring;, c = 6.94 & Aring;. EDS confirmed the K element proportion increased from 0.06% to 2.14%. Pyrolysis experiments demonstrated that BC-K significantly optimized gas composition: for cotton stalks, CO2 content in pyrolysis gas decreased by 21.67%, while H2 and CO yields markedly increased; for polypropylene plastic, the pyrolysis gas composition shifted distinctly toward a hydrogen-rich composition; for corn starch, H2 and CO yields were also effectively enhanced. DFT calculations revealed the modification mechanism at the atomic level: K+ doping reduced the charge to + 0.25e (compared to Ca2+, +0.37e), while CO32- carried a charge of -0.45e. This enhanced electron cloud delocalization and significantly increased the density of states near the Fermi level. The coupling between the p orbitals of oxygen, the s orbitals of potassium, and the p orbitals of carbon is strengthened, thereby enhancing the performance of the catalyst. This study provides a new technical pathway and theoretical basis for the resource utilization of waste biomass and the efficient energy conversion of carbon-hydrogen solid waste.

Metallo-covalent organic frameworks (M-COFs) integrate organometallic and polymer principles, with salen-type ligands offering a robust means to stabilize transition metal (TM) centers. Dual metallosalen based ligand systems (TM2C22H18N4O2) have been investigated using density functional theory (DFT) to elucidate how TM choice influences framework properties. The findings revealed that Sc, Ti, Y, and Zr atoms in these complexes exhibit pronounced out-of-plane distortions relative to the C22H18N4O2 ligand plane. Regarding thermodynamic stability, the Ti2C22H18N4O2 and ZrC22H18N4O2 complexes show superior performance compared to neighboring complexes. However, the complexes containing Ti2, Cr2, Cu2, Zr2, Mo2, and Cd2 display less chemical stability relative to adjacent systems. Additionally, Mulliken charge analysis shows that Sc2C22H18N4O2 and Y2C22H18N4O2 carry significantly higher charges than the remainder of the series, suggesting enhanced electron modulation capabilities. These insights guide the rational design of salen-based M-COFs for high-performance electrocatalysis.

Benefits from the pyrazole ligands 4, 5-bis(pyrazole-1-yl) phthalic acid (C14N4O4H10), Co(NO)3·6H2O and Cu(NO)3·3H2O, four 3D supramolecular compounds ([Co(C14N4O4H9)2(H2O)2]∙2H2O (1), [Co2(C16N4O4H13)2(CH3COO)2]·C2H5OH (2), [Co(C14N4O4H9)(CH3COO)(C3H7OH)]·C3H7OH (3)) and [Cu2(C14N4O4H9)2(H2O)2]∙6H2O (4) were synthesized. Among them, the ethanol solution in compound 2 can be used as both solvent and reactant to form 4, 5-bis(pyrazole-1-yl) ethyl phthalate with 4, 5-bis (pyrazole-1-yl) phthalate (C16N4O4H14), which is directly synthesized by a one-step method. This study reports a one-step synthesis method for ester compounds, opening new avenues for the synthesis of ester compounds using solvent. Compounds 1–4 were doped with graphite to prepare a series of composite carbon paste electrodes, and their electrochemical sensing performance toward uric acid, dopamine, and acetaminophen was investigated. Among the four compounds, the modified electrode prepared using 1 and 4 exhibited excellent electrochemical sensing performance, high sensitivity, exceptional selectivity, and remarkable recyclability for detecting uric acid (pH = 7), dopamine (pH = 4) and acetaminophen (pH = 7) in PBS solution.

The 6061 aluminum alloy (AA6061) is widely used in aerospace, automotive manufacturing, and electronic devices due to its excellent properties. As an advanced process integrating the advantages of casting and forging, semi-solid processing (SSP) technology can significantly improve the comprehensive properties of aluminum alloys. To address the difficulty in accurately predicting the deformation behavior of AA6061 in the two-phase region, isothermal compression tests were performed using a thermal simulator (THERCMESTOR-Z/100). Tests were conducted at strain rates of 0.01 to 10 s−1 and temperatures of 580 °C to 610 °C, and the corresponding stress-strain data were obtained. Meanwhile, combined with the analysis of the microstructure evolution law of AA6061, it was found that the presence of liquid phase plays a crucial role in the rheological behavior of the alloy in the two-phase region. Based on this, this paper constructed a modified Arrhenius model and a Fields–Backofen (FB) model by introducing a liquid volume fraction adjustment factor. The model accuracy was evaluated by the Average Absolute Relative Error (AARE) and coefficient of determination (R2). The results show that the AARE of the strain-compensated Arrhenius model was 16.1 pct, and the AARE of the Arrhenius model considering liquid volume fraction modification was 7.0 pct, with the prediction accuracy improved by 56.5 pct. The AARE of the FB model considering liquid-phase modification was 5.7 pct, which was the model with the highest prediction accuracy in this study, because the FB model further considered the piecewise characteristics of the stress-strain curve. The modified constitutive models significantly enhance the ability to describe the complex rheological behavior under non-equilibrium solidification conditions, providing theoretical support for the optimization of semi-solid processing windows and numerical simulation.