Search papers, labs, and topics across Lattice.
100 papers published across 7 labs.
SFG spectroscopy reveals molecular-level insights into solid-liquid interfaces, uncovering how probing depth varies with experimental setup and surface properties.
IDEAgent achieves a staggering 3.89x improvement in generating diverse and high-quality research ideas compared to existing methods.
TriGlue reveals a groundbreaking approach to molecular glue design that effectively combines interface estimation and ternary complex generation, paving the way for targeted protein degradation.
Anomalous malaria transmission patterns reveal that high-burden regions like Tamale may not be the most frequent hotspots, challenging traditional surveillance approaches.
Axel Becke's journey reveals how serendipity and timing can amplify scientific brilliance, shaping the very foundations of computational chemistry.
IDEAgent achieves a staggering 3.89x improvement in generating diverse and high-quality research ideas compared to existing methods.
TriGlue reveals a groundbreaking approach to molecular glue design that effectively combines interface estimation and ternary complex generation, paving the way for targeted protein degradation.
Anomalous malaria transmission patterns reveal that high-burden regions like Tamale may not be the most frequent hotspots, challenging traditional surveillance approaches.
Axel Becke's journey reveals how serendipity and timing can amplify scientific brilliance, shaping the very foundations of computational chemistry.
OpenGEM26 not only surpasses QM9 in conformational diversity but also enables a graph neural network that achieves unprecedented accuracy in predicting molecular behaviors.
AREX achieves superior performance in deep research tasks by recursively refining answers through a novel self-improvement mechanism that outpaces traditional search methods.
Encoding conformational ensembles into a single thermodynamically informed embedding can dramatically enhance cyclic-peptide property prediction accuracy.
ZF2ST achieves strong testing power for detecting distributional differences while maintaining rigorous statistical validity, outperforming traditional methods.
Neural networks can achieve percent-level accuracy in solving complex quantum field equations, outperforming traditional methods in robustness and stability.
HierarchicalDAEW achieves unprecedented accuracy in predicting spatial gene expression while providing robust uncertainty estimates, making it a game-changer for clinical applications in transcriptomics.
Automated machine learning models can achieve over 94% accuracy in characterizing charge states of quantum dots, paving the way for scalable quantum computing solutions.
Achieving high reconstruction fidelity with substantial data compression, GWCS redefines the efficiency of graph signal processing in scientific machine learning.
New independent set constructions push the Shannon capacity lower bounds for odd cycles, revealing the surprising power of LLMs in combinatorial discovery.
Cycle-consistency regularization boosts predictive accuracy to near perfection while enabling real-time parameter recovery in tokamak edge plasma simulations.
Transforming STI analytics, this framework enables LLMs to generate insights that are not only richer but also rigorously validated against scientific standards.
Hallucinations in chemical reasoning models coexist with correct answers, revealing a complex relationship that challenges our understanding of model reliability.
After 2022, scientists are not just collaborating more; they’re venturing into intellectually distant fields, reshaping the landscape of scientific inquiry.
Improved margin computation for STV elections could make risk-limiting audits both more effective and feasible, enhancing electoral integrity.
Long-range intermolecular interactions can generate a rich variety of hybrid states in molecular cavity QED, fundamentally altering our understanding of polaritonic behavior.
Unresolved satellite features in platinum's photoelectron spectrum reveal critical insights into its electronic structure and the role of relativistic effects.
Accurate calculations reveal that nitrogen's dispersion coefficients are critical for enhancing the precision of gas thermometry experiments.
Supported models reveal that causal processes can reach diverse eventual states from any starting point, challenging traditional views on logic programming semantics.
ASTRA-Net achieves a remarkable mean Dice score of 0.8927 for airway segmentation in drug-induced sleep endoscopy, even with minimal real annotations.
M$^3$-Gen reveals how multimodal integration can generate interpretable gene expression profiles, bridging the gap between clinical imaging and molecular data.
RD-SCL achieves superior performance in acoustic impedance imaging with only 56.5k parameters, revolutionizing the efficiency of few-shot learning in this domain.
The short-range interaction of nitrogen is shown to be pivotal for accurate metrological applications, with implications for understanding temperature-dependent properties.
A new loss function based on Itô's formula significantly improves the learning of committor functions, enhancing rare event simulations in molecular dynamics.
Correcting cumulant contributions in natural orbital functionals leads to exact energy predictions for bond dissociation, bridging a critical gap in RDMFT accuracy.
The theoretical estimates of nitrogen's polarizability reveal discrepancies with experimental data, opening avenues for refining metrological standards.
Achieving chemical accuracy in quantum phase estimation with a linearly scaling variational circuit could redefine practical applications on NISQ hardware.
PG-KINN outperforms legacy approaches by leveraging a Petrov-Galerkin framework, achieving robust solutions for complex PDEs with improved accuracy and interpretability.
Quantum autoencoders can match classical anomaly detection performance while fitting seamlessly into the resource-constrained environments of future collider experiments.
PIER outperforms standard retrieval methods by ensuring that environmental predictions are consistent with physical dynamics, leading to more accurate modeling of complex systems.
Broad classes of functions can now be analyzed without the failure phenomena previously observed, reshaping our understanding of convergence in random function contexts.
Gate twirling emerges as the most effective method for preserving quantum kernel geometry, yet paradoxically yields the lowest alignment with target labels.
Training neural surrogates without labeled data can achieve remarkable accuracy in predicting complex thermo-fluid fields, slashing model development costs.
OLEDLM can generate novel OLED candidates that meet stringent optoelectronic criteria, revolutionizing the search in a vast chemical space.
The unlearned skeleton outperforms learned methods at equal mesh resolution, challenging the assumption that learning always enhances performance in fluid dynamics.
Prioritizing candidate biomedical annotations with a novel framework using bioKGs boosts classifier accuracy and efficiency in expert curation.
DKMD isolates directional shifts in distributions while being robust to outliers and efficient enough to handle millions of samples in seconds.
AAMFM achieves unprecedented accuracy in functional antibody design by effectively pairing antibody sequences with antigen contexts, setting a new benchmark in the field.
Local causal structure learning can be achieved with high accuracy and efficiency even in the presence of latent variables and selection bias, challenging the need for global causal discovery methods.
TSFMs can forecast heart rate variability from consumer wearables with unprecedented accuracy, outperforming traditional methods without any fine-tuning.
HA-RFM achieves up to 100 times error reduction in high-dimensional elliptic PDEs by intelligently selecting features based on their structural relevance.
Automated symbolic regression can rediscover known HEP functions while achieving high-quality fits, streamlining data analysis in particle physics.
FMRP-LEAN transforms clinical biomarker workflows by integrating AI with a robust, HIPAA-compliant architecture that enhances operational efficiency and transparency.
PRIME-SVR achieves unprecedented reconstruction quality in fetal brain imaging, enabling the first isotropic T2 maps at previously inaccessible echo times.
DQAOA-GPT slashes computational costs while maintaining solution quality, outperforming traditional methods in tackling complex combinatorial optimization problems.
MOF-Sleuth transforms CIF auditing by linking chemical evidence directly to language model explanations, achieving unprecedented accuracy and clarity in diagnosing structural errors.
Embedding scattering mechanisms into complex-valued networks can significantly enhance the accuracy and consistency of PolSAR image classification.
SIINR not only super-resolves clinical dMRI images but also quantifies uncertainty, offering a game-changing approach for neuroimaging analysis.
SFG spectroscopy reveals molecular-level insights into solid-liquid interfaces, uncovering how probing depth varies with experimental setup and surface properties.
Achieving 93.8% accuracy in predicting molecular structures from multimodal spectroscopic data could revolutionize how chemists approach organic synthesis and analysis.
Collective electronic entanglement can be achieved without the typical O(1/N) dilution penalty, unlocking new possibilities for scalable quantum technologies.
Non-linear strain fields can dramatically alter mechanophore activation, challenging traditional views on reaction kinetics in mechanochemistry.
Varying the deposition rate can finely tune the domain sizes in organic thin films, unlocking new design strategies for organic electronic devices.
Neural networks can classify complex link topologies with 97% accuracy by leveraging the writhe density matrix, even in varying conditions.
In sparse autoregressive settings, CEDAR uncovers causal relationships with remarkable efficiency, requiring only $O(d^2)$ conditional independence tests after initial screening.
PhysCoRe outperforms traditional methods by accurately predicting deformable object dynamics while adapting to new materials and guiding exploration through confidence-based feedback.
RAINBOT\textsuperscript{TM} achieves autonomous liquid handling at a fraction of the cost of commercial systems, revolutionizing accessibility in laboratory automation.
RDMA-based communication in STORM achieves unprecedented scaling efficiency for Monte Carlo simulations, reducing overhead and enabling high-fidelity astrophysical modeling at scale.
Achieving chemical accuracy with just 4% of the configuration subspace could revolutionize how we approach quantum simulations of complex molecular systems.
AI-native NPUs can achieve competitive performance in scientific computing by addressing precision and memory bottlenecks through tailored workload mappings.
The model reveals that physical relationships in materials science are more accessible through controlled transformations than through static representations, challenging assumptions about how LLMs encode scientific knowledge.
Achieving over 10% improvement in molecular classification accuracy with a multi-modal approach reveals the untapped potential of combining diverse signal representations in nanopore sensing.
A new Monte Carlo estimator for Gaussian shape overlap achieves unprecedented accuracy and efficiency, cutting sampling needs by 94% while delivering reliable uncertainty estimates.
Remarkably, even incomplete or ill-conditioned descriptors can yield accurate atomic reconstructions, challenging the conventional wisdom about the necessity of high-dimensional feature sets.
Reliable assignments of core binding energies reveal the intricate electronic structure of solid-state amino acids, bridging a critical gap in our understanding of their chemistry.
Spin contamination in TDDFT can be resolved with a kernel reconstruction scheme that ensures internal consistency, eliminating artifact states and enabling a unified treatment of target states.
Rem3Di redefines molecular representation by enabling the differentiation of enantiomers without relying on classical 2D fingerprints, achieving state-of-the-art results in property prediction.
A denoiser trained solely on classical statistics can accurately reproduce quantum Boltzmann distributions without retraining, revealing a profound connection between generative modeling and quantum fluctuations.
Accurate branching ratios in nonadiabatic molecular dynamics can be achieved by rescaling nuclear momentum in a novel way during electronic collapses.
Uncovering hidden patient groups in breast cancer data could transform how clinicians approach personalized treatment strategies.
Super-resolved atmospheric data can now be trusted to respect fundamental physics, enhancing extreme event detection accuracy.
Target-aligned input reparameterization slashes prediction errors for thermodynamic properties in supercritical combustion by up to 14.5 times, revolutionizing neural network efficiency.
A new hybrid model predicts electrolyte behaviors for 9,296 systems, overcoming the limitations of traditional models that require extensive experimental data.
DBMol transforms small molecule design by effectively optimizing binding affinities while ensuring chemical validity, setting a new standard for de novo molecular generation.
Exact boundary enforcement in PINNs isn't enough; the choice of distance function can make or break the accuracy of solutions to elliptic Dirichlet problems.
Bayesian linear regression informed by physics can dramatically enhance the reliability of energy consumption forecasts for electric trucks, outperforming traditional methods.
ATLAS achieves over 500-fold efficiency in sampling amorphous materials while maintaining less than 0.2% free energy error, revolutionizing the approach to material design.
LLMol outperforms traditional methods by directly optimizing molecular generation through verifiable rewards, achieving remarkable efficiency and success rates in complex design tasks.
SCGP achieves superior glucose forecasting by explicitly conditioning on individual patient data, outperforming traditional methods that fail to personalize predictions.
ABOPD achieves a remarkable 0.42 Å reduction in RMSD for antibody CDR design, setting a new standard for structural fidelity in protein generation.
Mechanical signals can provide critical early warnings for lithium-ion battery thermal runaway, achieving a lead time that outpaces existing methods by nearly 70%.
Identifying probabilistic structures from vanishing binomials reveals interpretable language patterns without traditional parameter estimation.
Residual HOS2 boosts Jaccard similarity from 0.382 to 0.460, revealing critical insights into the challenges of cooperative gene regulatory recovery.
Models that excel in prediction often falter in mechanistic reasoning, revealing hidden flaws in their logic and understanding of cellular context.
Memory retrieval in dense networks hinges on initial states, with the exact full-retrieval threshold for the LSE model now clearly defined.
AI agents struggle to surpass 50% accuracy in genomic surveillance tasks, revealing significant gaps in their analytical capabilities.
Current LLMs achieve only a 12.30% success rate in generating executable scientific code, highlighting a significant gap in their capabilities.
Trademark data uncovers the real-world impact of AI technologies on product innovation, revealing insights that patents alone cannot provide.
A novel Company World Model outperforms traditional biotech departments in driving AI-native drug development success.
Noise-induced attacks on VQE can amplify errors by up to 8.84 times, highlighting critical vulnerabilities in quantum computing pipelines.
Shifting the critical frequency in EIS measurements can drastically cut down acquisition time while maintaining accuracy, revolutionizing how we approach electrochemical characterization.
The one-shot $G_0W_0$ kernel can misestimate superconductivity onset temperatures by orders of magnitude, revealing the critical need for self-consistent treatments in polar semiconductors.
FlareEUV outperforms traditional methods in predicting solar EUV irradiance, revealing critical insights into flare dynamics that could enhance space weather forecasting.
Shallow seismic activities can obscure the critical accumulation of deep tectonic strain, revealing a hidden risk in reservoir-fault systems.
Current records of stadium crowd noise are misleadingly based on single-point measurements, but a new framework reveals the need for spatially distributed assessments to ensure accuracy.
GeoDES achieves unprecedented accuracy in storm structure synthesis, outperforming existing models and redefining weather prediction capabilities.
Achieving high-fidelity inference in gravitational lensing, this method reveals intricate details of galaxy mass distributions that were previously obscured by noise.
