Search papers, labs, and topics across Lattice.
100 papers published across 3 labs.
Efficiently reducing metal artifacts in CBCT without manual intervention, our method sets a new standard for reconstruction accuracy and practicality in medical imaging.
Current AI research agents miss critical metacognitive checks, leading to pervasive failures across diverse scientific tasks.
Unlocking the potential of multimodal AI to interpret complex scientific images could revolutionize how we access and understand experimental data.
Personalization in AI co-scientists could be the key to unlocking novel research insights that generic systems overlook.
Faraday, the AI Scientist, not only outperforms leading models in replication tasks but also adopts a more rigorous scientific approach, hinting at a new era of AI-driven research.
Current AI research agents miss critical metacognitive checks, leading to pervasive failures across diverse scientific tasks.
Unlocking the potential of multimodal AI to interpret complex scientific images could revolutionize how we access and understand experimental data.
Personalization in AI co-scientists could be the key to unlocking novel research insights that generic systems overlook.
Faraday, the AI Scientist, not only outperforms leading models in replication tasks but also adopts a more rigorous scientific approach, hinting at a new era of AI-driven research.
Neural ODEs can revolutionize transformer thermal forecasting by integrating physics directly into the modeling process, yielding robust predictions with less data.
Aggregating quantum information across nodes without losing fidelity is now possible, even for non-Gaussian codes, thanks to a novel measurement-based approach.
Inter-residue contact distances reveal the hidden molecular motions driving protein dynamics, bridging the gap between experimental spectroscopy and molecular simulations.
Nested sampling reveals how Cu(100) oxidation transitions through complex surface reconstructions and defect states, challenging traditional thermodynamic modeling approaches.
A single qubit can exponentially reduce measurement requirements for learning classical signals, revolutionizing sensing tasks and applications.
A single learned functional can accurately predict diverse thermodynamic behaviors without relying on traditional training targets.
SbCD achieves unprecedented accuracy in generating complete crystallographic structures by effectively modeling symmetry-breaking dynamics.
SORT shifts the focus from brittle term selection to adaptive basis design, enabling more robust and accurate equation discovery from noisy data.
The leading error in MYULA is driven by the active trace of the Moreau envelope, not the global curvature, allowing for faster convergence rates in high-dimensional settings.
UltraIR outperforms traditional machine-learning approaches in IR spectroscopy, achieving high accuracy with minimal labeled data across diverse chemical analysis tasks.
MK-TGAN outperforms traditional generative models by integrating biological knowledge, producing synthetic transcriptomic data that is both realistic and useful for research.
Integrating simulated data with real-world measurements reveals hidden patterns and discrepancies that traditional data-driven methods miss.
EGRL achieves an impressive 8.6% improvement in AUROC for cold-start RNA-protein interactions, showcasing its potential to revolutionize RPI prediction.
Noise doesn't have to derail your search for fractional differential equations—Weak-Pareto recovers accurate structures even under significant measurement noise.
Fine-tuned Normalizing Flows can dramatically reduce simulation costs while improving accuracy in high-energy physics experiments.
A physics-informed autoencoder achieves sub-1.1% error in DRT estimation across multiple datasets, revealing latent representations that track operational changes and degradation in solid oxide cells.
Legacy bioinformatics code can be transformed into efficient Rust implementations, slashing size by 80x and build time by 10x while boosting performance over threefold.
VLSR's localize-then-reason approach boosts throughput by 9.6X, revolutionizing how we analyze molecular properties from images.
Recursive self-improvement in quantitative trading research leads to a groundbreaking Sharpe ratio of +2.50, showcasing the potential for autonomous systems to enhance investment strategies over time.
AI is driving the cost of mathematical outputs to zero, but the true value lies in the human journey of understanding that remains underfunded and at risk.
Efficiently reducing metal artifacts in CBCT without manual intervention, our method sets a new standard for reconstruction accuracy and practicality in medical imaging.
Princigram achieves a new standard in scientific diagram generation, ensuring physical accuracy through a structured reasoning framework that traditional models lack.
Bypassing ground-truth dependencies, this framework achieves unprecedented accuracy in white matter tractography through a novel hybrid reinforcement and supervised learning approach.
Class-geometry supervision transforms how we approach sample-efficient open-world detection, yielding significant gains in novel-class insertion and unknown recall.
Nonadiabatic corrections can dramatically alter the conductivity and Drude weight in electron-ion systems, challenging conventional models of energy transfer.
AQE11D achieves unprecedented accuracy at extreme tolerances, but at a staggering computational cost compared to the more efficient DE formula.
Achieving a 5.1x speedup in a legacy weather simulation while ensuring scientific validity reveals the critical role of validation in AI-assisted GPU porting.
LAAB transforms performance reporting for mathematical libraries, ensuring that every benchmark is traceable and relevant to real-world scientific applications.
Electrostatic embedding can dramatically enhance binding free energy predictions for specific targets, but its effectiveness varies significantly across different systems.
Achieving a fivefold speedup in solvation free energy calculations while maintaining accuracy could revolutionize computational chemistry practices.
Intern-S2-Preview-397B not only excels in multimodal scientific reasoning but also enhances biological instruction performance without altering its foundational architecture.
Lifecycle-wide perception is crucial for evidence-grounded scientific discovery, allowing OmniScientist to outperform traditional AI systems in research quality and depth.
ScreenShot can predict effective drug combinations from minimal patient data, achieving superior accuracy while cutting experimental costs by two-thirds.
O3 outperforms traditional methods at low oracle budgets, but FK-steering and DPO take the lead as budgets increase—showing that context matters in model guidance.
The advective Fisher-Rao metric not only optimizes probability measure paths but also outperforms conventional methods in fitting probability densities.
Symmetry increase in Equivariant Neural Networks can be systematically reduced, enhancing their expressivity for symmetric inputs without sacrificing performance.
A novel Bayesian approach reveals that robust damage detection in structural health monitoring can be achieved even in the presence of unmeasured environmental variability.
Half of the variance in phototransistor gain comes from differences between production runs, fundamentally challenging traditional predictive models.
Operators can be learned more efficiently with kernel methods that outperform deep learning models while maintaining high predictive accuracy.
A global CNN model for AGB estimation can be effectively calibrated with minimal local data, achieving superior accuracy and addressing the challenges of sparse field inventories.
RECAST achieves up to 92% error reduction in coarse-grid PDE simulations, proving that machine learning can drastically enhance numerical accuracy without increasing computational costs.
MLLMs excel at reasoning over diagrams but falter in parsing and editing them, revealing a significant gap that needs addressing.
Achieving a p-value of 0.0078 in the RxRx1 study highlights a powerful new approach to disentangling treatment effects from complex structured outcomes.
Ground states of quantum systems can now be computed efficiently for up to 8100 qubits using a novel foundation model that redefines traditional approaches.
Deformable convolution outperforms traditional methods, achieving a remarkable 20.79 dB PSNR and 0.9623 Dice score in nanophotonic absorber design.
New orthogonal multiwavelets outperform traditional filters in image compression and denoising, achieving superior visual quality metrics.
Achieving 99.5% citation validity and 96.4% figure editability, Spark-to-Paper transforms how research papers can be generated with unprecedented reliability and efficiency.
Invertible transformations can simplify complex binary-linear operators to a canonical form, revealing deeper connections in the structure of quadratic APN maps.
OpenMP Offloading can deliver up to 4x performance gains in multi-GPU setups, outperforming traditional single-GPU implementations across major architectures.
Memory effects in electronic friction can drastically alter energy dissipation dynamics, leading to unexpected increases in vibrational energy loss during scattering events.
Machine-learning potentials can achieve DFT-level accuracy in simulating ionic-liquid impacts while being four orders of magnitude faster than traditional methods.
Last-layer embeddings in protein language models often underperform, with shallow layers proving more effective for specific datasets like deep mutational scans.
Controlling the diagonalization subspace is crucial for accurate benchmarking of quantum sampling methods, as classical random sampling can outperform SQD under certain conditions.
Ions in water exhibit multiple rattling frequencies that reveal their solvation dynamics and could influence selective ion transport and binding behaviors.
The Breit contribution to correlation energy can reach 4% for heavy elements like xenon, revealing critical insights into relativistic effects in quantum chemistry.
Personalized feedback-driven recommendations can boost literature discovery effectiveness by over 10% in aligning with user preferences.
Open-weight model adoption in scientific research is not just a trend toward transparency; it's a geopolitical shift, with Chinese researchers leading the charge.
Achieving critical-point preservation in vector-field compression at GPU speeds up to 640 times faster than traditional CPU methods could revolutionize data handling in scientific simulations.
Achieving high throughput in quantum decoding at both room and cryogenic temperatures could redefine the efficiency of fault-tolerant quantum computing systems.
Most protein backbone generation models may be recycling known structures rather than creating truly novel folds, as revealed by the new Domain Retrieval Rate metric.
Incompletely labeled PROTAC databases can still yield powerful predictors for protein degradation, achieving state-of-the-art performance without extensive experimental labels.
Aquaporins are not just water channels; they are multifunctional gates that regulate ions and adapt to physiological challenges across life forms.
Phase space electronic structure theory could revolutionize our understanding of spin-related chemical phenomena, offering insights that traditional methods struggle to provide.
The second-derivative correction to FANPT leads to a more reliable continuation strategy for nonlinear wavefunction equations, reducing parameter deviations by leveraging the overlap Hessian.
Exact entropy expressions for finite classical many-body systems reveal surprising ensemble discrepancies that challenge traditional thermodynamic equivalence.
Spin rotatory strength reveals unexpected dynamics in chiral-induced spin selectivity, challenging conventional views on chirality in quantum materials.
Achieving chemical accuracy with triple-$\zeta$ basis sets in transcorrelated calculations could redefine efficiency benchmarks in computational chemistry.
Achieving line-center fitting errors below 0.00001 cm$^{-1}$ sets a new standard for precision in molecular spectroscopy.
LLMs engage with flawed scientific premises 95% of the time, revealing a critical gap in their epistemic reliability for scientific applications.
Rational numbers are just the tip of the iceberg—Reversible Chemical Reaction Networks can compute a much richer set of real numbers than previously understood.
Rationale supervision can boost performance on complex scientific problems but may backfire on simpler ones, revealing critical insights into AI training dynamics.
Bayesian Neural Networks outperform other methods in handling uncertainty in genomics, revealing critical insights for high-stakes biological predictions.
CI tests can significantly influence causal discovery outcomes, yet many methods struggle with robustness and scalability in complex biomedical datasets.
ChemWorld allows researchers to isolate the impact of hidden chemical laws on agent behavior, enabling unprecedented control and replayability in autonomous chemistry experiments.
No single normalization method excels universally in spectral clustering; effectiveness hinges on network density and community structure.
MR-MoL reveals how GNN-derived attributions can transform molecular property predictions by making the reasoning process transparent and interpretable.
Autoresearch agents can waste compute resolving the same issues repeatedly, but targeted interventions can dramatically enhance their efficiency and performance.
Ignoring geometric nonlinearity in thermal device design can lead to critical performance errors, particularly as operating temperatures rise.
Accurately resolving sharp discontinuities in hyperbolic conservation laws is now feasible without the pitfalls of prior assumptions or artificial smoothing.
Cross-path reasoning reveals untapped research ideas, outperforming traditional methods and rivaling human insights in scholarly evolution.
AI can generate novel mathematical insights, as evidenced by tightening the Grothendieck constant bounds significantly.
A new domain-specific language empowers researchers to automate hybrid quantum-classical problem-solving, making it easier to identify when quantum advantages can be leveraged.
Switching to a backup controller when uncertainty spikes can drastically reduce safety violations in robotic chemistry tasks.
Achieving a mere $1.6\times$ speedup from a theoretical $4\times$ peak illustrates the critical need for holistic optimization in matrix computations.
Transient X-ray absorption spectra reveal that nearly symmetric proton-sharing configurations dramatically enhance absorption, reshaping our understanding of molecular dynamics in acetylacetone.
Machine-learned proxies can revolutionize how we map free energy landscapes in ionic colloidal systems, revealing that subtle changes in surface potentials can dramatically shift crystal structure preferences.
Achieving over 60-fold signal enhancements in NMR for unmodified amino acids could revolutionize the detection of low-concentration metabolites.
Short-term precipitation nowcasts can transform into actionable risk indicators, empowering agricultural supply chains to make proactive decisions in the face of climate variability.
ERRLESS reveals that incorporating Bayesian principles into symbolic regression can yield more interpretable and accurate models, outperforming traditional methods in uncertainty quantification.
Achieving polynomial sample complexity for approximating solutions to complex elliptic PDEs could revolutionize the efficiency of numerical methods in stochastic modeling.
Jointly predicting mutations in lung cancer can significantly boost diagnostic accuracy, but only for the right combinations of genes.
Automating rank determination in tensor-based neural networks could revolutionize their application in complex physics simulations.
A physics-guided approach in VeinCast reveals that integrating structured atmospheric relations can significantly enhance the accuracy of medium-range weather forecasts.
Achieving a 2.10% reconstruction error with a method that eliminates preliminary region selection could revolutionize 3-D electromagnetic imaging.
The Beta-Variational Autoencoder uncovers hidden lunar anomalies, revealing both geological wonders and human artifacts with unprecedented accuracy.
A novel multimodal architecture allows a machine to generate accurate lunar geological interpretations while effectively citing scientific chronologies.
