Browse Articles

Improving LLM performance on olympiad-level mathematics through cognitive decomposition

Abstract Large Language Models often struggle with Olympiad-level mathematics due to logical errors in multi-step proofs. This paper evaluates the impact of cognitive decomposition on reasoning by comparing four architectures: two single-agent systems (Chain-of-Thought and Metacognitive) and two multi-agent systems (Sequential Refinement and a parallel Council of Specialists). Using the OlympiadBench dataset, our results demonstrate that both the Council of Specialists (65.45% accuracy) and the Metacognitive Single-Agent (64.12%) significantly outperform the CoT Single-Agent (55.15%). Critically, the performance difference between the multi-agent Council and the Metacognitive Single-Agent is not statistically significant. This establishes that explicit cognitive decomposition, specifically the structured separation of strategy, pattern seeking, and constraint analysis, is the primary driver of reasoning improvements, rather than multi-agent parallelism. Domain-level analysis reveals that the Sequential architecture remains competitive in Combinatorics. A leave-one-out ablation study further reveals that the Council’s performance relies on distinct specialist roles: the Pattern Seeker is critical for Number Theory and Geometry, while the Constraint Analyst is critical for Algebra. We recommend the Metacognitive Single-Agent as the optimal architecture, as it delivers the robust reasoning benefits of a multi-agent council within a single prompt at less than half the computational cost.

N-MIX: an in silico framework for predicting ADAM10-mediated substrate cleavage sites through structural and spatial analysis of protease–substrate interactions

Disentangled graph–text collaborative filtering with prototypical cold-start for scientific article recommendation

Bio-stimulants improve tomato growth by regulating the rhizosphere microbiome involved in phosphorus and nitrogen cycling

Nationwide trends in outpatient rates for ophthalmic surgeries in Japan before and after the COVID-19 pandemic: an analysis of the NDB open data

Changes in heritability and shared environmentality of educational attainment across twentieth-century Norway

Electrochemical CO2 capture from engine exhaust using NaCl electrolyte: System modeling, optimization, and economic assessment

Metal–organic framework glass enables durable sodium-ion storage for hard carbon negative electrodes

Optimization of excavation method for large-section metro station tunnel: a case study

Abstract The Baijusi Station Tunnel of Chongqing Rail Transit Line 18, excavated in moderately weathered sandstone, was used as a case study to optimize the excavation method for a large-section metro station tunnel. The double side drift (DSD) and optimized three-bench (OTB) methods were first compared using Hoek–Brown-based numerical simulations. The results showed comparable tunnel stability for the two methods, while the OTB method offered advantages in construction efficiency and working space. A trial section was then excavated to verify the OTB method, with the central pillar dismantled as the upper bench advanced. Field monitoring showed that deformation and support responses remained within acceptable ranges, and the spatial distribution of microseismic events reflected micro-crack evolution broadly consistent with the numerically predicted rupture surfaces. Under the investigated geological, loading, and construction conditions, the central pillar did not play a decisive role in controlling overall tunnel stability. The OTB method was subsequently applied to the entire tunnel under the same monitoring scheme, and the monitored deformation, support stress, and microseismic responses remained within acceptable ranges. These results indicate that, with appropriate numerical assessment, trial verification, and field monitoring, eliminating the central pillar can be feasible under comparable conditions, achieving a balance between tunnel stability and construction efficiency.

Contrasting patterns of alpine biodiversity across mountains and taxa worldwide

Abstract The alpine biome, located at higher elevations of mountains worldwide, supports unique biodiversity and provides important ecosystem contributions to people. Despite the growing recognition of mountain biodiversity in international policy frameworks, substantial gaps remain in our understanding of how alpine biodiversity varies across mountain systems, undermining estimates of its conservation value and consequently effective conservation strategies. Here, we curate a dataset on alpine biodiversity, incorporating expert-validated data on species’ elevational ranges for vascular plants, mammals, birds, and reptiles across 32 mountain ranges worldwide. We show that alpine biodiversity hotspots are concentrated in Neotropical regions, while most temperate regions represent coldspots with lower species richness. These patterns persist whether considering species with broad elevational ranges or only those strictly confined to the alpine zone. Unlike the classical latitudinal gradient of biodiversity, alpine richness patterns show no consistent relationship with latitude, highlighting the importance of regional history, landscape structure, and biogeographical processes.

Liquid time constant based neuromorphic active inference for resilient control and health aware battery management in hybrid AC/DC microgrids

DNA transposon expansion drives genome plasticity in Diutina catenulata

TWEAK/FN14 inhibition synergizes with oncogene-directed tyrosine kinase inhibitors to overcome resistance across multiple driver contexts

Transient distortions of the South Atlantic Anomaly radiation environments driven by electric fields

Abstract Energetic electrons in Earth’s inner radiation belt pose significant hazards to spacecraft systems, with the strongest radiation in low-Earth orbit (LEO) mostly confined to the South Atlantic Anomaly (SAA) region. Once considered stable, the inner belt is now understood to exhibit significant variability. Using data from the low-Earth-orbit Macau Science Satellite-1 mission, we report transient distortions of the SAA radiation environments, observationally characterized by enhanced fluxes of energetic electrons outside the traditional SAA radiation region, appearing either attached to or detached from its boundary. We show that these distortions can be explained by large-scale electric-field perturbations that adiabatically alter the electron mirror heights, which can be further modulated by ultra-low-frequency waves. Test-particle simulations successfully reproduce the observational features and provide crucial constraints on properties of the associated electric fields. These findings reveal a distinct manifestation of inner-belt variability, extending the electron radiation risks beyond the expected boundaries of the SAA radiation environments.

A hypofunctional PREX1 variant (p.Y191C) leads to neurodevelopmental abnormalities and epilepsy by attenuating RAC1 signaling

G-quadruplex homeostasis is a determinant of PARP inhibitor toxicity in BRCA2-deficient cells

Abstract Inhibition of PARPs is a key strategy to treat tumours with defects in homologous recombination (HR), including those with mutations in the tumour suppressor gene BRCA2 . PARP inhibitors generate replication stress, creating a dependence on HR to repair the resulting DNA damage. However, the DNA lesions generated upon PARP inhibition that impede replication fork progression and trigger a requirement for BRCA2 in cell survival are poorly defined. Here, we demonstrate that elevated levels of G-quadruplex (G4) DNA structures is a determinant of genome instability and PARP inhibitor toxicity, while suppressing these structures results in PARP inhibitor resistance. The HUWE1-associated stress response protein HAPSTR1 and BRCA2 function in parallel pathways to PARP1/PARP2 to suppress G4 levels during S-phase. Mechanistically, PARP1/PARP2 disruption in HAPSTR1 or BRCA2-deficient cells leads to G4-replication conflicts, ssDNA gaps, replication-associated DNA damage and genome instability. HAPSTR1 turnover is regulated through HUWE1-dependent proteasome degradation. As such, HUWE1 disruption results in elevated HAPSTR1 and suppression of elevated G4 levels in BRCA2-deficient cells, resulting in PARP inhibitor resistance. Together, these data identify G4 structures as a determinant of PARP inhibitor toxicity, while the HAPSTR1/HUWE1 axis is essential to suppress these structures and confer PARP inhibitor resistance.

Biochemical and growth responses of mungbean to actinobacterial inoculation under water stress

Author Correction: Economical biogas direct methanation to pipeline grade natural gas via structured Ni based inverse catalyst

Kat5 deficiency in alveolar type II cells licenses STAT6-driven glycolytic reprogramming and pulmonary fibrosis

Cryo-EM structure of the naked mole-rat ribosome reveals a stabilized split 28S rRNA

Abstract The naked mole-rat (Heterocephalus glaber) is a long-lived mammal with resistance to cancer and hypoxia, suggesting the evolution of robust proteostasis networks. The ribosome, central for protein synthesis, is key to cellular stress responses and has an unusual feature: the 28S rRNA split; however, the details of its organization remain unknown. Here, we present high-resolution cryo-EM structures of the naked mole-rat 80S ribosome in four states of the elongation cycle. The structures reveal a conserved overall architecture and rRNA modification landscape compared to other mammals, and provide an atomic-level view of the distinct break in the 28S rRNA. This cleavage event, located in the D6 expansion segment, is structurally stabilized by a network of interactions with surrounding ribosomal proteins, maintaining the integrity of the large subunit. Our comparative analysis revealed that this compensatory network preserves a canonical architecture that is nearly indistinguishable from intact mouse and human ribosomes. These findings resolve the structural basis of this distinct cleavage, showing that it is a stable, integrated feature whose function is likely linked to more subtle regulatory mechanisms, rather than inducing major structural rearrangements.