Search for: All records

The rapid advancement of large language models (LLMs) has opened new possibilities for AI for good applications. As LLMs increasingly mediate online communication, their potential to foster empathy and constructive dialogue becomes an important frontier for responsible AI research. This work explores whether LLMs can serve not only as moderators that detect harmful content, but as mediators capable of understanding and de-escalating online conflicts. Our framework decomposes mediation into two subtasks: judgment, where an LLM evaluates the fairness and emotional dynamics of a conversation, and steering, where it generates empathetic, de-escalatory messages to guide participants toward resolution. To assess mediation quality, we construct a large Reddit-based dataset and propose a multi-stage evaluation pipeline combining principle-based scoring, user simulation, and human comparison. Experiments show that API-based models outperform open-source counterparts in both reasoning and intervention alignment when doing mediation. Our findings highlight both the promise and limitations of current LLMs as emerging agents for online social mediation. 

Abstract Freshwater discharge from ice sheets induces surface atmospheric cooling and subsurface ocean warming, which are associated with negative and positive feedbacks respectively. However, uncertainties persist regarding these feedbacks’ relative strength and combined effect. Here we assess associated feedbacks in a coupled ice sheet-climate model, and show that for the Antarctic Ice Sheet the positive feedback dominates in moderate future warming scenarios and in the early stage of ice sheet retreat, but is overwhelmed by the negative feedback in intensive warming scenarios when the West Antarctic Ice Sheet undergoes catastrophic collapse. The Atlantic Meridional Overturning Circulation is affected by freshwater discharge from both the Greenland and the Antarctic ice sheets and, as an interhemispheric teleconnection bridge, exacerbates the opposing ice sheet’s retreat via the Bipolar Seesaw. These results highlight the crucial role of ice sheet-climate interactions via freshwater flux in future ice sheet retreat and associated sea-level rise. 

De novo design of molecules with targeted properties represents a new frontier in molecule development. Despite enormous progress, two main challenges remain: (i) generating novel molecules conditioned on targeted, continuous property values; (ii) obtaining molecules with property values beyond the range in the training data. To tackle these challenges, we propose a reinforced regressional and conditional generative adversarial network (RRCGAN) to generate chemically valid molecules with targeted HOMO–LUMO energy gap (ΔEH–L) as a proof-of-concept study. As validated by density functional theory (DFT) calculation, 75% of the generated molecules have a relative error (RE) of <20% of the targeted ΔEH–L values. To bias the generation toward the ΔEH–L values beyond the range of the original training molecules, transfer learning was applied to iteratively retrain the RRCGAN model. After just two iterations, the mean ΔEH–L of the generated molecules increases to 8.7 eV from the mean value of 5.9 eV shown in the initial training dataset. Qualitative and quantitative analyses reveal that the model has successfully captured the underlying structure–property relationship, which agrees well with the established physical and chemical rules. These results present a trustworthy, purely data-driven methodology for the highly efficient generation of novel molecules with different targeted properties. 

We report the effects of screening capacity, surface roughness, and interfacial epitaxy of the bottom electrodes on the polarization switching, domain wall (DW) roughness, and ferroelectric Curie temperature (T_C) of PbZr_0.2Ti_0.8O₃ (PZT)-based free-standing membranes. Singe crystalline 10−50 nm (001) PZT and PZT/La_0.67Sr_0.33MnO₃ (LSMO) membranes are prepared on Au, correlated oxide LSMO, and two-dimensional (2D) semiconductor MoS₂ base layers. Switching the polarization of PZT yields nonvolatile current modulation in the MoS₂ channel at room temperature, with an on/off ratio of up to 2 × 10⁵ and no apparent decay for more than 3 days. Piezoresponse force microscopy studies show that the coercive field E_c for the PZT membranes varies from 0.75 to 3.0 MV cm^-1 on different base layers and exhibits strong polarization asymmetry. The PZT/LSMO membranes exhibit significantly smaller E_c, with the samples transferred on LSMO showing symmetric E_c of about −0.26/+0.28 MV cm^-1, smaller than that of epitaxial PZT films. The DW roughness exponent ζ points to 2D random bond disorder dominated DW roughening (ζ = 0.31) at room temperature. Upon thermal quench at progressively higher temperatures, ζ values for PZT membranes on Au and LSMO approach the theoretical value for 1D random bond disorder (ζ = 2/3), while samples on MoS2 exhibits thermal roughening (ζ = 1/2). The PZT membranes on Au, LSMO, and MoS₂ show T_C of about 763 ± 12, 725 ± 25, and 588 ± 12 °C, respectively, well exceeding the bulk value. Our study reveals the complex interplay between the electrostatic and mechanical boundary conditions in determining ferroelectricity in free-standing PZT membranes, providing important material parameters for the functional design of PZT-based flexible nanoelectronics. 

Gyroid structure, a nature inspired cellular architecture, is under extensive exploration recently due to its structure continuity, uniform stress distribution under compression, and stable collapse mechanism during deformation. However, when combining with a functional gradient, the Gyroid structure can perform much different mechanical behavior from its homogeneous counterpart. Herein, bottom-up computational modeling is performed to investigate the mechanics of functional gradient nano-gyroid structure made of copper (Cu). Our work reveals that its mechanical properties degrade with a density that is much slower than those of homogeneous gyroid structure. The scaling of yield strength [Formula: see text] to the relative density [Formula: see text] for the functional gradient gyroid structure is in the factor of 1.5. Moreover, the layer-by-layer collapsing mechanism yields significantly better mechanical energy absorption ability. This study not only leads to insightful understanding of the deformation mechanisms in nonuniform gyroid structures but also promotes the development of the functional gradient cellular materials.