We introduce MoS (Mixture of States), a novel fusion paradigm for multimodal diffusion models that merges modalities using flexible, state-based interactions. The core of MoS is a learnable, token-wise router that creates denoising timestep- and input-dependent interactions between modalities' hidden states, precisely aligning token-level features with the diffusion trajectory. This router sparsely selects the top-k hidden states and is trained with an ε-greedy strategy, efficiently selecting contextual features with minimal learnable parameters and negligible computational overhead. We validate our design with text-to-image generation (MoS-Image) and editing (MoS-Editing), which achieve state-of-the-art results. With only 3B to 5B parameters, our models match or surpass counterparts up to 4× larger. These findings establish MoS as a flexible and compute-efficient paradigm for scaling multimodal diffusion models.

Chain-of-thought (CoT) reasoning has emerged as a powerful tool for multimodal large language models in video understanding. However, its necessity and advantages over direct answering remain underexplored. In this paper, we first demonstrate that for RL-trained video models, direct answering often matches or even surpasses CoT performance, despite CoT producing step-by-step analyses at a higher compute cost. Motivated by this, we propose VideoAutoThink, a video understanding framework that adopts a ``reason-when-necessary'' strategy. During training, our approach follows a Thinking Once, Answering Twice paradigm: the model first generates an initial answer, then performs reasoning, and finally outputs a reviewed answer. Both answers are supervised with verifiable rewards. During inference, the model uses the confidence score of the initial answer to determine whether to proceed with reasoning. Across video QA and grounding benchmarks, VideoAutoThink achieves state-of-the-art accuracy with significantly improved efficiency, reducing the average response length by ~3.3x, e.g., from 144 to just 44 tokens. Moreover, we observe low activation of thinking on perception-oriented tasks, but higher activation on reasoning-intensive tasks. This suggests that explicit language-based reasoning is generally beneficial but not always necessary.

Intrinsic image memorability is a stable, predictable property of photographs, hitherto studied mainly as a perceptual phenomenon. Using computer-vision estimates of memorability, we link large-scale user-generated images to consumer evaluation and valuation. Across 154,945 Yelp photos, consumable depictions are more memorable than indoor and outdoor scenes, and balanced people–object composition predicts higher memorability. At the business level (N = 26,769), higher average photo memorability is associated with higher star ratings only for high-visibility businesses (top tertile of review count), consistent with a trust-and-risk boundary condition. In an online experiment (N = 268), presenting same-category foods in high- versus low-memorability images is associated with a 32.8% increase in maximum willingness to pay (from 18.96 to 25.18 CNY; paired d = 0.65), and predicted memorability tracked stimulus-level recognition. Together, these results identify actionable visual correlates of memorability and show that memorability pays off mainly for established, high-visibility businesses.

The rapid advancement of Large Multimodal Models (LMMs) for 2D images and videos has motivated extending these models to understand 3D scenes, aiming for human-like visual-spatial intelligence. Nevertheless, achieving deep spatial understanding comparable to human capabilities poses significant challenges in model encoding and data acquisition. Existing methods frequently depend on external depth sensors for geometry capture or utilize off-the-shelf algorithms for pre-constructing 3D maps, thereby limiting their scalability, especially with prevalent monocular video inputs and for time-sensitive applications. In this work, we introduce VLM-3R, a unified framework for Vision-Language Models (VLMs) that incorporates 3D Reconstructive instruction tuning. VLM-3R processes monocular video frames by employing a geometry encoder to derive implicit 3D tokens that represent spatial understanding. Leveraging our Spatial-Visual-View Fusion and over 200K curated 3D reconstructive instruction tuning question-answer (QA) pairs, VLM-3R effectively aligns real-world spatial context with language instructions. This enables monocular 3D spatial assistance and embodied reasoning. To facilitate the evaluation of temporal reasoning, we introduce the Vision-Spatial-Temporal Intelligence benchmark, featuring over 138.6K QA pairs across five distinct tasks focused on evolving spatial relationships. Extensive experiments demonstrate that our model, VLM-3R, not only facilitates robust visual-spatial reasoning but also enables the understanding of temporal 3D context changes, excelling in both accuracy and scalability.

High-efficiency formamidinium lead iodide (FAPbI3)-based perovskite solar cells (PSCs) typically involve annealing in humid air during the fabrication process of perovskite films. However, the combined effects of humidity and relatively high temperature often result in the uncontrollable formation of a detrimental PbI2 phase in the perovskite films. As a result, the annealing process of perovskite films is highly sensitive to the relative humidity fluctuations of the environment. Under solar illumination, the undesired PbI2 tends to decompose, accelerating the degradation of perovskite materials and severely compromising the light stability of PSCs. This issue is particularly critical for the buried interface and bulk of the perovskite films, as these regions absorb the majority of the incident light. Pre-treatment and post-treatment strategies are generally confined to address the PbI2 issues at the buried interface and on the surface of the perovskite films, respectively. However, effectively addressing the effects of excess PbI2 at buried interface and grain boundaries within bulk in a single step remains challenging. In this study, we propose an intermediate-treatment strategy using phthalylglycyl chloride (PTC), which involves treating the wet films with PTC prior to annealing during the formation process of the perovskite films. This approach protects the grain boundaries of polycrystalline perovskite films in advance, effectively preventing moisture-induced degradation of the perovskites and thus significantly broadening the relative humidity window of annealing process. Our results demonstrate that this strategy can successfully suppress the formation of PbI2 at the grain boundaries and buried interface of perovskite films, thereby eliminating the PbI2-induced degradation pathways. Our strategy significantly reduces the sensitivity to humidity fluctuations during annealing for fabricating stable PSCs, ensuring more consistent fabrication of stable PSCs. Consequently, the resulting PSCs achieve a champion power conversion efficiency of 26.1% and demonstrate excellent light stability.

The realm of drug development and supramolecular chemistry has been revolutionized by the emergence of macrocyclic compounds, which have emerged as formidable contenders. Glecaprevir is an 18-membered macrocycle used for the treatment of chronic hepatitis C virus infection. The inherent structural flexibility of glecaprevir's side chain enables it to adopt diverse solution conformations, thereby influencing its binding modes with biological proteins and unlocking new potentials for the treatment of other diseases. The present study delved into the solution structural characteristics of glecaprevir through a combination of experimental UV-Vis and electronic circular dichroism spectra, complemented by theoretical calculations. Our findings reveal that glecaprevir consistently exhibits spectral characteristics in both protic and aprotic solvents due to its stable conformational distribution. DFT-assisted structural analyses were conducted to explore the weak noncovalent interactions within the stable conformations of glecaprevir. The molecular docking analysis conducted on a diverse range of conformers of glecaprevir revealed its exceptional selectivity against proteins associated with breast cancer, colorectal cancer, lung cancer, and lymphoma. These remarkable findings suggest that glecaprevir holds immense potential as a lead compound for drug design targeting these specific types of cancers. Furthermore, the conformationally adapted mechanism is indicated by the alteration of intramolecular hydrogen bonds subsequent to molecular docking.

Cyclic peptide–lenalidomide conjugated molecules were prepared and a comprehensive investigation into their antitumor activity was performed.

Suppressing side reactions led to the formation of rational Dion–Jacobson 2D/3D surface heterostructures, achieving an optimal PCE of 23.05% (certified 22.58%) for 1.68 eV perovskite solar cells.

Interfacial molecules have been demonstrated to improve the photovoltaic performance of perovskite solar cells (PSCs). However, the effect is influenced by the targeted substrate and, in particular, by its bond with the interfacial molecule. A weaker bonding of the interfacial molecule with the substrate usually implies a stronger bonding with the perovskite that could lead to uncontrollable insertion of the interfacial molecule into the perovskite bulk, resulting in device degradation. Here we select bis(2-aminoethyl) ether (BAE) as the interfacial molecule between the perovskite and the electron transport layer (ETL) in n-i-p PSCs and develop a strategy to harmonize the strength of the bilateral bonds of BAE. In particular, we manipulate the electronic structure of the ETL with doping to increase the strength of the BAE-ETL bond. This thereby results in a weakening of the BAE-perovskite bond. The harmonization in bilateral bonds of the interfacial molecule leads to PSCs with an efficiency surpassing 26.5% (certified as 26.31%) and improved stability. The interfaces in perovskite solar cells are critical to the device performance. Li et al. tune the bond strength of the interfacial molecule with the perovskite and the electron transport layer, increasing the power conversion efficiency of the cells.

As the initial synthesized colloidal quantum dots (CQDs) are generally capped with insulating ligands, ligand exchange strategies are essential in the fabrication of CQD films for solar cells, which can regulate the surface chemical states of CQDs to make them more suitable for thin-film optoelectronic devices. However, uncontrollable surface adsorption of water molecules during the ligand exchange process introduces new defect sites, thereby impairing the resultant device performance, which attracts more efforts devoted to it but remains a puzzle. Here, we develop a solvent-engineering-assisted ligand exchange strategy to revamp the surface adsorption, improve the exchange efficiency, and modulate the surface chemistry for the environmentally friendly lead-free silver bismuth disulfide (AgBiS 2 ) CQDs. The optimized AgBiS 2 CQD solar cells deliver an outstanding champion power conversion efficiency (PCE) of up to 8.95 % and improved long-term stability. Our strategy is less environment-dependent and can produce solar cells with negligible performance variance for several batches across several months. Our work demonstrates the critical role of solvents for ligand exchange in the surface chemistry of CQDs and the realization of high-performance photovoltaic devices in a highly reproducible manner.

Perovskite light-emitting diodes (PeLEDs) are the next promising display technologies because of their high color purity and wide color gamut, while two classical emitter forms, i.e., polycrystalline domains and quantum dots, are encountering bottlenecks. Weak carrier confinement of large polycrystalline domains leads to inadequate radiative recombination, and surface ligands on quantum dots are the main annihilation sites for injected carriers. Here, pinpointing these issues, we screened out an amphoteric agent, namely, 2-(2-aminobenzoyl)benzoic acid (2-BA), to precisely control the in situ growth of FAPbI(3) (FA: formamidine) nanodomains with enhanced space confinement, preferred crystal orientation, and passivated trap states on the transport-layer substrate. The amphoteric 2-BA performs bidentate chelating functions on the formation of ultrasmall perovskite colloids (<1 nm) in the precursor, resulting in a smoother FAPbI(3) emitting layer. Based on monodispersed and homogeneous nanodomain films, a near-infrared PeLED device with a champion efficiency of >22% plus enhanced T-80 operational stability was achieved. The proposed perovskite nanodomain film tends to be a mainstream emitter toward the performance breakthrough of PeLED devices covering visible wavelengths beyond infrared.

The exploration of novel anticancer compounds based on natural cyclopeptides has emerged as a pivotal paradigm in the contemporary advancement of macrocyclic pharmaceuticals. Phakellistatin 13 is a cycloheptapeptide derived from the brown snubby sponge and exhibits remarkable antitumor activity. In this study, we have designed and synthesized a series of chiral cyclopeptides incorporating the rigid isoindolinone moiety at various sites within the natural cycloheptapeptide Phakellistatin 13, with the aim of investigating conformationally constrained cyclopeptides as potential antitumor agents. Cyclopeptide 3, comprising alternating l-/d-amino acid residues, exhibited promising antihepatocellular carcinoma effects. Detailed biological experiments have revealed that Phakellistatin 13 analogs effectively inhibit the proliferation of tumor cells and induce apoptosis and autophagy, while also causing cell cycle arrest through the modulation of the p53 and mitogen-activated protein kinase (MAPK) signaling pathway. This study not only provides valuable insights into chemical structural modifications but also contributes to a deeper understanding of the biological mechanisms underlying the development of natural cyclopeptide-based drugs.

Surface ligand chemistry is vital to control the synthesis, diminish surface defects, and improve the electronic coupling of quantum dots (QDs) toward emerging applications in optoelectronic devices. Here, we successfully develop highly homogeneous and dispersed AgBiS2 QDs, focus on the control of interdot spacing, and substitute the long-chain ligands with ammonium iodide in solution. This results in improved electronic coupling of AgBiS2 QDs with excellent surface passivation, which greatly facilitates carrier transport within the QD films. Based on the stable AgBiS2 QD dispersion with the optimal ligand state, a homogeneous and densely packed QD film is prepared by a facile one-step coating process, delivering a champion power conversion efficiency of approximately 8% in the QD solar cells with outstanding shelf life stability. The proposed surface engineering strategy holds the potential to become a universal preprocessing step in the realm of high-performance QD optoelectronic devices.

Macrocyclic molecules have emerged as significant contenders in drug development and supramolecular chemistry. Paritaprevir is a 15-membered macrocycle approved for the treatment of chronic Hepatitis C virus infection. The structural flexibility inherent in paritaprevir enables the presentation of diverse conformational features in solutions, thereby influencing its binding pattern with proteins. In this paper, we investigated the structural characteristics of paritaprevir in protic solvents (methanol and trifluoroethanol) as well as an aprotic solvent (acetonitrile) by using experimental electronic circular dichroism spectra combined with density functional theory calculations. Our study reveals that paritaprevir can adopt different molecular conformations due to its flexible carbon chain and the solvent environment. Conformer B is particularly stable in polar solvents because of the planar hydrophobic interactions between the phenanthridine and methylpyrazine groups. Density functional theory calculations-assisted structural analyses further disclosed the weak noncovalent interactions within the stable conformations of paritaprevir. Molecular docking of paritaprevir on different cancer proteins showed that the macrocycle was highly selective against lymphoma, endometrial carcinoma, colorectal cancer, and breast cancer. This suggests that paritaprevir could potentially serve as a promising lead compound for drug design targeting these specific types of cancers. The comparison between the low energy conformation and pharmacophore conformation (docking conformation) indicates that conformer B is the primary pharmacophore conformation of paritaprevir, providing important macrocyclic structural characteristics for future macrocyclic drug development.

Additive engineering has emerged as one of the most promising strategies to improve the performance of perovskite solar cells (PSCs). Among additives, perovskite nanocrystals (NCs) have a similar chemical composition and matched lattice structure with the perovskite matrix, which can effectively enhance the efficiency and stability of PSCs. However, relevant studies remain limited, and most of them focus on bromide-involved perovskite NCs, which may undergo dissolution and ion exchange within the FAPbI3 host, potentially resulting in an enlarged band gap. In this work, we employ butylamine-capped CsPbI3 NCs (BPNCs) as additives in PSCs, which can be well maintained and serve as seeds for regulating the crystallization and growth of perovskite films. The resultant perovskite film exhibits larger domain sizes and fewer grain boundaries without compromising the band gap. Moreover, BPNCs can alleviate lattice strain and reduce defect densities within the active layer. The PSCs incorporating BPNCs show a champion power conversion efficiency (PCE) of up to 25.41 %, well over both Control of 22.09 % and oleic acid/oleylamine capped CsPbI3 NC (PNC)-based devices of 23.11 %. This work illustrates the key role of nanosized seed surfaces in achieving high-performance photovoltaic devices.

The target affinity of cyclopeptides to MDM2 protein can be improved by adjusting the molecular electrostatic potential and volume.

The techniques for 3D indoor scene capturing are widely used, but the meshes produced leave much to be desired. In this paper, we propose "RoomDreamer", which leverages powerful natural language to synthesize a new room with a different style. Unlike existing image synthesis methods, our work addresses the challenge of synthesizing both geometry and texture aligned to the input scene structure and prompt simultaneously. The key insight is that a scene should be treated as a whole, taking into account both scene texture and geometry. The proposed framework consists of two significant components: Geometry Guided Diffusion and Mesh Optimization. Geometry Guided Diffusion for 3D Scene guarantees the consistency of the scene style by applying the 2D prior to the entire scene simultaneously. Mesh Optimization improves the geometry and texture jointly and eliminates the artifacts in the scanned scene. To validate the proposed method, real indoor scenes scanned with smartphones are used for extensive experiments, through which the effectiveness of our method is demonstrated.

The limited conductivity of existing transparent conducting oxide (TCO) greatly restricts the further performance improvement of perovskite solar cells (PSCs), especially for large-area devices. Herein, buried-metal-grid tin-doped indium oxide (BMG ITO) electrodes are developed to minimize the power loss caused by the undesirable high sheet resistance of TCOs. By burying 140-nm-thick metal grids into ITO using a photolithography technique, the sheet resistance of ITO is reduced from 15.0 to 2.7 Ω sq-1 . The metal step of BMG over ITO has a huge impact on the charge carrier transport in PSCs. The PSCs using BMG ITO with a low metal step deliver power conversion efficiencies (PCEs) significantly better than that of their counterparts with higher metal steps. Moreover, compared with the pristine ITO-based PSCs, the BMG ITO-based PSCs show a smaller PCE decrease when scaling up the active area of devices. The parallel-connected large-area PSCs with an active area of 102.8 mm2 reach a PCE of 22.5%. The BMG ITO electrodes are also compatible with the fabrication of inverted-structure PSCs and organic solar cells. The work demonstrates the great efficacy of improving the conductivity of TCO by BMG and opens up a promising avenue for constructing highly efficient large-area PSCs.

Monitoring extent and severity is vital in the ulcerative colitis (UC) follow-up, however, current assessment is complex and low cost-effectiveness. We aimed to develop a routine blood-based clinical decision support tool, Jin’s model, to investigate the extent and severity of UC. The multicentre retrospective cohort study recruited 975 adult UC inpatients and sub-grouped into training, internal validation and external validation set. Model was developed by logistics regression for the extent via Montreal classification and for the severity via Mayo score, Truelove and Witts score (TWS), Mayo endoscopic score (MES) and Degree of Ulcerative colitis Burden of Luminal Inflammation (DUBLIN) score. In Montreal classification, left-sided and extensive versus proctitis model achieved area under the receiver operating characteristic curve (AUROC) of 0.78 and 0.81 retrospectively. For severity, Mayo score model, TWS model, MES model and DUBLIN score model achieved an AUROC of 0.81, 0.70, 0.74 and 0.70 retrospectively. The models also were evaluated with satisfactory calibration and clinical unity. Jin’s model was free with open access at http://jinmodel.com:3000/ . Jin’s model is a noninvasive, convenient, and efficient approach to assess the extent and severity of UC.