Search papers, labs, and topics across Lattice.
100 papers published across 7 labs.
Autonomous research agents can now produce publication-grade manuscripts in frontier physics, achieving substantive findings while navigating complex literature landscapes.
A multi-agent framework that combines prior knowledge with data-driven analysis can significantly enhance causal discovery in complex, high-dimensional datasets.
Most enzyme specificity models fail to outperform basic sequence alignment methods, highlighting a critical gap in predictive capabilities.
Automation can't eliminate decision uncertainty without expanding access to physical records, revealing critical limits in autonomous scientific control.
Kramers' classic model is reinterpreted to show that the trajectory's history is crucial for accurately mapping chemical kinetics, challenging conventional separation methods.
Most enzyme specificity models fail to outperform basic sequence alignment methods, highlighting a critical gap in predictive capabilities.
Automation can't eliminate decision uncertainty without expanding access to physical records, revealing critical limits in autonomous scientific control.
Kramers' classic model is reinterpreted to show that the trajectory's history is crucial for accurately mapping chemical kinetics, challenging conventional separation methods.
Two-dimensional coherent spectroscopy reveals distinct contributions to spectral linewidths, outperforming traditional methods in clarity and precision.
EPMF is NP-hard even for the simplest case, revealing deep computational challenges in low-rank matrix approximations.
Optimizing multithreading in event generators can drastically cut energy costs, paving the way for sustainable high-energy physics research.
QSPADE revolutionizes quantum anomaly detection by enabling efficient, noise-resistant anomaly scoring without the computational burden of traditional PCA methods.
Injection-driven, weakly supervised training can achieve reliable Real-Bogus classification without human labels, providing calibrated uncertainties and robust performance under class contamination.
Bridging the gap between coarse atmospheric models and local PM$_{2.5}$ variations, this framework achieves a 40x super-resolution without relying on temporal data.
Target-Guided Selective Reweighting PINN achieves superior parameter recovery in inverse problems without sacrificing field accuracy, even amidst challenging source-target mismatches.
Pathway Activity Autoencoders reveal that integrating multi-omics data can significantly enhance cancer risk stratification and survival predictions while maintaining interpretability.
FUSE resolves complex parameter degeneracies in exoplanet orbital estimation, outperforming state-of-the-art methods and closely matching ground-truth MCMC results.
Navigating high-dimensional Time-Derivatives spaces reveals rich, interpretable insights from complex dynamical systems without the need for dimensionality reduction.
PDEFlow automates the entire process from user input to solver-free inference, enabling rapid experimentation with complex differential equations.
By harnessing the power of symmetries in structured data, geometric causal models reveal new pathways for causal inference that traditional methods overlook.
A physics-informed variational autoencoder captures complex spatial correlations in bridge damage assessment, achieving 77.2% accuracy in uncertainty quantification.
Soft labeling transforms DNA read classification, achieving a 2.56× reduction in error and paving the way for more accurate cell-type deconvolution across diverse biological samples.
Achieving unprecedented structural continuity and geological realism in velocity model reconstruction, even with sparse data, could transform subsurface imaging practices.
Predicting breast cancer treatment response just got a major upgrade, with a new framework that outperforms traditional models by effectively modeling temporal imaging data.
A novel U-Net framework significantly boosts extreme precipitation forecast accuracy, turning previously negligible predictions into operationally valuable insights.
MARLIN reveals that reliable molecular structure elucidation is possible without any prior knowledge of the molecular formula, challenging the assumptions of current methods in mass spectrometry analysis.
QSM-derived continuous biomass references can enhance AGB estimation accuracy by correcting edge-effect uncertainties, especially in small field plots.
A new simulation-free method for reconstructing population dynamics outperforms traditional approaches by seamlessly handling large gaps in data.
Conventional benchmarks miss critical structural failures in quantum circuit design, but HamQASBench exposes hidden pitfalls that could derail QAS efforts.
LLMs may excel at strategic thinking, but they lag behind specialized models in practical synthesis planning, highlighting a critical gap in current AI capabilities.
Solving the Supersingular Isogeny Problem using Gr\"obner basis techniques can be dramatically faster than conventional approaches, reshaping the efficiency landscape of cryptographic applications.
Secure key agreement is possible even in noisy environments, thanks to twin optical PUFs that can withstand fabrication variability.
A new typology reveals that the coherence of chart-image pairs in scientific communication can significantly impact how experts and non-experts interpret data.
Retinal graph phenotypes can prioritize systemic pathways in diabetic retinopathy, revealing critical mediators like glycaemic–renal interactions that traditional methods overlook.
DriftST achieves one-step generative inference of gene expression from H&E images, outperforming traditional methods that struggle with inter-gene dependencies and resolution flexibility.
A new gauge transformation reveals how to eliminate numerical instabilities in quantum dynamics simulations, unlocking previously unreachable coupling strengths.
The new approach resolves the inverse problem in TAP reactor analysis, yielding significantly improved reaction-rate reconstructions, particularly for nonlinear kinetics.
Cavity quantum electrodynamics can now be harnessed to fine-tune chemical reactions with near-exact accuracy, revealing new pathways in polaritonic chemistry.
A single rod defect in quadrupole mass filters can create unexpected transmission ridges and nonlinear resonances, challenging conventional designs.
A novel automated workflow enables the efficient design of ionic electrolytes, yielding a dataset that uncovers the intricate interplay between solvent composition and battery performance.
Ternary nucleobase mixtures crystallize inefficiently due to kinetic competition, but preformed seeds can dramatically accelerate and direct the crystallization process.
Charge conditioning in ML force fields can drastically enhance predictive accuracy while maintaining computational efficiency, achieving remarkable reductions in error metrics with minimal data.
Unigram-LM outperforms BPE by segmenting chemical structures into 29-41% more tokens, revealing the critical impact of tokenizer choice on model performance.
Integrating GNSS-derived Zenith Wet Delay into weather models boosts severe precipitation forecasts by nearly 9%.
Uncovering novel black hole solutions, this method reveals metrics with trapped interiors that challenge existing paradigms in general relativity.
Achieving high-precision quantum chemical calculations on a consumer GPU could democratize access to advanced computational methods in chemistry.
Over six million commits from CERN reveal the hidden impact of institutional contributions to open source software.
ResearchStudio-Idea transforms the ideation process by systematically grounding proposals in literature and identifying unresolved research bottlenecks, leading to more robust and traceable research directions.
None of the 30 LLM agents evaluated in CausalGame demonstrated reliable causal thinking, revealing a critical gap in AI's ability to perform scientific reasoning.
Quantum mechanics and environmental factors impose strict speed limits on biomolecular processes, revealing that traditional estimates may be fundamentally flawed.
Symbolic regression could unlock new efficiencies in kinetic energy functional construction, challenging the dominance of deep neural networks in DFT applications.
A single agent can dramatically reduce experimental costs and iterations by intelligently balancing high- and low-cost measurements based on predicted uncertainties.
Spectral diffusion enables DynaMode to predict protein dynamics with unprecedented accuracy, distinguishing between slow and fast motions effectively.
A groundbreaking approach to Gilbreath's conjecture reveals new insights into prime number patterns and opens doors to advanced applications in cybersecurity and randomness testing.
Achieving four orders of magnitude higher thermal stability, this Ising machine redefines the potential for practical applications in combinatorial optimization.
The itinerant oscillator model reveals that traditional assumptions in EIS interpretation can obscure critical insights about electrolyte dynamics and timescale distributions.
Trotterization can break symmetries in quantum algorithms, but a novel approach using operator kirigami preserves these symmetries, enhancing accuracy in quantum simulations.
Switching from Adam to SOAP or SOAP-Muon can dramatically accelerate training and boost accuracy in machine learning interatomic potentials.
The Spin-MInt algorithm's symplecticity is proven for any number of electronic states, a significant leap in the accuracy of nonadiabatic quantum dynamics simulations.
HRPA(D) and SOPPA(CCSD) outperform traditional methods in accuracy while revealing that RPA excels for aromatic molecules due to its unique electronic excitation energy overestimation.
Memory constraints fundamentally alter the landscape of stabilizer state testing and learning, revealing a surprising collapse of the established complexity separation.
Q-GAIN transforms cold-atom research by seamlessly combining machine learning with physics-informed analysis, enabling rapid and accurate feature detection in complex datasets.
The framework reveals distinct contributions of chemical and structural factors to aqueous solubility, enhancing both predictive accuracy and interpretability in drug discovery models.
Early detection of Alzheimer's is revolutionized by identifying critical biomarkers through advanced machine learning techniques, potentially transforming patient management.
Pre-training boosts CLMs' molecular structure awareness, but surprisingly, even randomly initialized models excel at encoding ring structures from the first layer.
Integrating biological interaction networks into Gaussian processes boosts minority-class performance in omics classification, outperforming traditional methods.
Predicting time increments with a compact Fourier Neural Operator leads to faster and equally accurate modeling of Rayleigh-Bénard convection.
Autonomous research agents can now produce publication-grade manuscripts in frontier physics, achieving substantive findings while navigating complex literature landscapes.
FitOne outperforms general-purpose LLMs by up to 10% on fitness certification exams, showcasing the power of domain-specific training in AI applications.
MolSight achieves unprecedented accuracy in molecular image reasoning by seamlessly integrating chemical topology with vision-language processing.
Large-scale structured academic visual data can transform image generation from mere aesthetic appeal to verifiable knowledge-grounded creation.
Achieving a mean absolute error of just 1.68 mm in 4D heart mesh reconstruction could revolutionize cardiac digital twin applications.
APEIRON revolutionizes TDAQ systems by seamlessly integrating hardware and software to enhance data processing in high energy physics experiments.
LLMs don't just capitulate to skepticism; they exhibit nuanced responses that can misrepresent their understanding of scientific consensus.
A multi-agent framework that combines prior knowledge with data-driven analysis can significantly enhance causal discovery in complex, high-dimensional datasets.
Coding agents can now reliably replicate scientific claims, ensuring that computational results are not just generated but thoroughly validated against original research.
A single structural edit can drastically impair LLM performance in molecular tasks, highlighting the fragility of their generalization capabilities.
HPG-Diff achieves remarkable compliance errors of just 0.87% in-distribution while drastically reducing floating material ratios, transforming topology optimization practices.
DiscoPER not only automates hypothesis generation but also self-analyzes its discoveries, revealing hidden patterns and expanding the search space in unprecedented ways.
A single model can simultaneously generate shape-aligned molecules and feasible synthesis plans, overcoming the traditional trade-offs in drug design.
GAIA achieves a 64% reduction in error for airfoil flow reconstruction, setting a new benchmark for operator learning in complex geometries.
Projected quantum kernels can drastically improve sample efficiency in Gaussian process optimization, outperforming traditional high-dimensional quantum kernels.
Separable graphs reveal a hidden structure in graphical models that could unify diverse independence frameworks and streamline model identification.
ILLUME+ reveals that a multi-faceted approach to explainability can uncover hidden molecular interactions in cancer treatment, outperforming traditional methods.
FPPF not only tackles the degeneracy issue in particle filters but also achieves superior performance in high-dimensional data assimilation tasks, redefining the landscape of filtering techniques.
Achieving an RPD greater than 2.0 with low overfitting, this machine learning approach revolutionizes rapid soil nutrient quantification, crucial for sustainable agriculture.
AI-generated molecules could harbor hidden dangers, but MolSafeEval reveals their safety risks through a comprehensive evaluation framework.
AgentODE reveals that even with limited data, mechanistic insights can be extracted from population-level statistics, challenging the reliance on individual-level data in rare disease modeling.
An automated model that predicts deadly outcomes of acute myocardial infarction outperforms traditional methods, potentially saving lives through faster diagnosis.
Label Influence Propagation reveals that dynamically adjusting label influences can significantly enhance multi-label node classification performance, outperforming existing methods.
Land-cover type trumps individual environmental factors as the key predictor of bird diversity in Sri Lanka, with urbanization favoring generalists at the expense of overall richness.
Achieving optimal scheduling in autonomous labs can cut experimental time significantly, even under complex hardware constraints.
A fully automated LLM framework expands chemical reaction classification from 68 to over 14,000 classes, achieving 97.7% accuracy on unseen data without human curation.
MuRFiV achieves unprecedented long-term prediction accuracy in spatiotemporal dynamics by merging finite-volume principles with deep learning, outperforming conventional neural networks.
BrainFIBRE reveals neurobiologically interpretable representations that outperform existing methods in predicting critical brain health markers from microstructural data.
AnF-DiffPET achieves unprecedented denoising performance, significantly enhancing low-dose PET imaging quality by addressing anatomical and frequency inconsistencies.
Traditional models misestimate admittance in nanocapacitors, but this new framework reveals critical insights into charge dynamics that could redefine our understanding of nanoscale energy storage.
AI2 methods can now be transformed from bespoke protocols into reusable workflows, significantly enhancing the efficiency of molecular simulations in complex chemical systems.
Heavier isotopes of $D_2$ reveal a striking reduction in scattering asymmetry due to longer dissociation times, reshaping our understanding of quantum coherence in electron interactions.
A new mechanism reveals that isolated donor-chiral bridge-acceptor complexes can achieve significant spin polarization without relying on spin-orbit coupling.
Graph-native reinforcement learning can boost hypothesis generation in materials science by achieving up to 65% better traceability than traditional models.
Fault-tolerant multi-gate teleportation can reduce entanglement costs from $n$ ebits to just 1, while effectively managing correlated errors in noisy networks.
Nodal-line structures in chemical dynamics are surprisingly robust under symmetry, but their stability is not guaranteed when that symmetry is broken.
Active-GRPO not only outperforms existing methods in molecular optimization but also redefines how models can adaptively balance imitation and self-discovery during training.
NNPs can achieve near-chemical accuracy in enzyme catalysis predictions with less than 1,000 system-specific data points, revolutionizing the efficiency of mechanistic studies.
