Search papers, labs, and topics across Lattice.
100 papers published across 4 labs.
Distributed GPU training slashes the time needed to train deep learning models for CFD, making accurate fluid simulation predictions accessible in a fraction of the time.
Multi-turn medical AI agents trained with RL tend to collapse into verbose, single-turn monologues, but a novel self-distillation method can restore multi-turn tool use and improve performance.
Get a sneak peek at the future of cluster science through the eyes of leading researchers tackling the dynamics of electrons in atomic and molecular nanoclusters.
PINNs get a wavelet makeover, adaptively focusing on high-magnitude source regions and leaving vanilla methods in the dust on PDEs with extreme loss imbalances.
Training on D3-Gym, a new dataset of real-world scientific tasks with verifiable environments, closes the gap between open-source and proprietary models on ScienceAgentBench by 7.8 points.
Multi-turn medical AI agents trained with RL tend to collapse into verbose, single-turn monologues, but a novel self-distillation method can restore multi-turn tool use and improve performance.
Get a sneak peek at the future of cluster science through the eyes of leading researchers tackling the dynamics of electrons in atomic and molecular nanoclusters.
PINNs get a wavelet makeover, adaptively focusing on high-magnitude source regions and leaving vanilla methods in the dust on PDEs with extreme loss imbalances.
Training on D3-Gym, a new dataset of real-world scientific tasks with verifiable environments, closes the gap between open-source and proprietary models on ScienceAgentBench by 7.8 points.
Unlocking interpretable clinical forecasting: StructGP recovers causal relationships and patient progression patterns directly from irregular EHR data, outperforming black-box methods in accuracy and uncertainty calibration.
Standard GNNs can't cut it for solving linear SDPs, but a carefully designed architecture that mimics first-order solver updates can learn to predict solutions and dramatically accelerate convergence.
You can accurately predict steel hardness from nanoindentation data with a tiny dataset and some clever physics-based data augmentation, even when traditional methods fail.
Forget computationally verifying stability – VibroML automatically *fixes* dynamically unstable crystal structures, opening the door to exploring previously inaccessible materials.
Physics-informed "grey-box" models aren't just more accurate for structural health monitoring, they can also be greener by reducing computational costs and carbon emissions.
Homomorphic encryption can make federated learning nearly as accurate as centralized training on sensitive healthcare data, but at a steep computational cost, while differential privacy offers a less expensive but accuracy-sacrificing alternative.
Even with high-quality data, systematic biases in stellar age inference can lead to stable but incorrect conclusions about the Milky Way's formation history, undermining the reliability of big data in Galactic archaeology.
Segmenting tiny brain arteries just got a whole lot better: a new loss function boosts Dice scores by up to 10% on these critical but challenging structures.
Current multimodal LLMs struggle to understand scientific spectra, but a new benchmark and data processing technique could change that.
Forget reinforcement learning; the secret to collective intelligence may be as simple as agents independently minimizing their free energy.
LLMs can accurately recall constraints while simultaneously violating them, with "knows-but-violates" rates ranging from 8% to 99%, revealing a fundamental flaw in multi-turn ideation.
LLMs reveal that research data is being reused far more often than previously thought, suggesting open science's impact is bigger than we realized.
ChatGPT for Clinicians, not human doctors, currently achieves the highest scores on a new benchmark of real-world clinical LLM tasks.
A lightweight CNN can achieve 97% accuracy in classifying mango leaf diseases, offering a practical solution for early disease detection in agriculture.
A new test split for DeepSpaceYoloDataset helps push the boundaries of automated astronomical object detection by providing a more diverse and challenging evaluation benchmark.
A single self-supervised model trained on millions of unlabeled brain MRI slices can generalize across diverse neuroimaging tasks, rivaling or exceeding specialized models, even with limited labeled data.
Forget bulky IR cameras: ThermoMesh's tiny thermoelectric mesh can pinpoint sparse heat sources with remarkable sensitivity thanks to clever material science.
Distributed GPU training slashes the time needed to train deep learning models for CFD, making accurate fluid simulation predictions accessible in a fraction of the time.
Salt's impact on proteins isn't just about ionic strength—it's a delicate dance between ions and hydration water, finally visualized through advanced simulation.
Reactive MLIPs can now accurately model electrochemical interfaces thanks to a differentiable layer that dynamically identifies molecular fragments and conserves charge within them.
Unlock direct biomarker detection in physiological fluids without dilution: flexoelectric resonance in silicon nanowires overcomes Debye screening limitations.
Long-lived molecular states, weakly coupled to chaotic short-range dynamics, offer a surprising route to stable quantum control in ultracold collisions.
Calculating X-ray absorption spectra for heavy element systems just got a whole lot faster, thanks to a new relativistic method that slashes computational costs without sacrificing accuracy.
Polymorph selection in metal-organic frameworks happens surprisingly early, starting at the pre-nucleation cluster stage.
By explicitly modeling uncertainty in hypergraph refinement, UHR-Net achieves more accurate segmentation of challenging lesions in medical images.
Achieve clinically relevant accuracy in dynamic bronchoscopy without breath-hold protocols by modeling patient-specific respiratory deformation within a Gaussian splatting framework.
Domain-specific scientific models, previously siloed from LLM agent systems, can now be orchestrated for complex reasoning tasks via the Eywa framework, unlocking performance gains on structured data.
AI research agents can now reliably trace method evolution topologies thanks to a new methodological evolution graph, Intern-Atlas, that captures structured relationships between research methods.
Machine learning can turn sparse simulation data into a complete phase diagram for collective motion models, revealing nuanced phase boundaries.
By jointly embedding spatial biology, histology, and clinical data, Haiku lets you ask "what if" questions about disease progression, revealing molecular shifts linked to clinical outcomes.
Half of pedestrian crashes outside intersections happen surprisingly close to them, suggesting intersection design flaws may have a larger impact than previously thought.
Accurately predicting Alzheimer's progression just got a major boost: PROMISE-AD uses longitudinal data and a Transformer-based survival framework to achieve state-of-the-art performance in forecasting conversion from cognitively normal to MCI and MCI to AD.
AutoML on plant electrical signals spots water stress with 92% accuracy, beating deep learning and offering farmers a faster route to optimized irrigation.
Gradient attribution in AI weather models offers a computationally validated, model-informed approach to reward allocation in participatory weather sensing, but beware: adversarial inputs can game the system.
Ditch the hash: training-free Hyper-Dimensional Fingerprints (HDF) unlock molecular representations with superior structural fidelity and property prediction compared to conventional methods, even at low dimensions.
Guaranteeing charge balance in generated amorphous materials is now possible without sacrificing accuracy or efficiency, thanks to AMGenC's novel approach.
Feature-level contrastive learning with dynamic masking unlocks superior performance on tabular remote sensing data, even when labels are scarce.
Forget prompt engineering – a structured methodology using LLM "helper agents" can measurably improve the efficiency and performance of LLM agents in complex scientific domains.
LLMs can prune noisy edges in EEG graphs, leading to more accurate and interpretable seizure detection.
Turns out, language models can reason about mechanical engineering problems, iteratively refining linkage designs by diagnosing failure modes and proposing grounded corrections, all without fine-tuning.
A generative model of human physiology not only beats existing clinical risk scores at predicting disease, but also accurately simulates the effects of clinical interventions, paving the way for personalized medicine.
Automating the translation of economic intuitions into executable computational experiments is now possible, potentially accelerating the pace of economic research.
Ditching text chunks for full document page images in medical RAG boosts QA accuracy by a full percentage point, proving that visual context matters.
Automated segmentation of radiological Peritoneal Cancer Index (rPCI) regions from CT scans is now feasible, potentially replacing invasive surgical assessment for peritoneal metastases.
Discovering new molecules and materials just got 10x cheaper, thanks to a hybrid AI method that blends generative models with physics-based search.
Recipes, like languages, exhibit universal statistical laws governing their structure, suggesting a deeper, shared cognitive basis for creative expression across cultures.
Unlock collaborative AI development in genomics without compromising patient privacy: this framework lets multiple institutions jointly train synthetic data generators on sensitive RNA-seq data using MPC and DP.
Achieve robust ionogram track separation, even under disturbed ionospheric conditions with unknown track numbers, by integrating physical models into fuzzy clustering.
Achieve superior CT-MRI cervical spine registration by adaptively fusing Mamba-based global context with Swin Transformer-based local detail.
Achieve high-fidelity CBCT reconstructions from ultra sparse-view data by decoupling geometry and texture in 3D Gaussian Splatting, enabling physically consistent residual detail compensation.
Seemingly strong segmentation models can fail at clinically critical tumor-vessel interfaces, highlighting the need for uncertainty-aware AI in pancreatic cancer staging.
Fusing dermoscopic images, clinical photos, and patient metadata with adaptive weighting dramatically improves skin lesion classification, even in imbalanced, real-world clinical datasets.
Continuous-depth transformers, augmented with physics-informed loss, can significantly improve short-term weather forecasting, suggesting a promising path for hybrid physics-aware AI models.
Skip the SCF convergence grind: a physically-constrained equivariant neural net slashes the number of iterations needed by up to 81% while also predicting accurate molecular properties in a single shot.
The familiar concept of the chemical bond may be a useful descriptor, but it's not the fundamental *cause* of molecular stability, challenging chemists' intuitive understanding of structure.
Carbonates blasted with radiation at Europa-like temperatures spontaneously generate and trap CO2 with the same spectral fingerprint as observed on the moon, finally explaining a major mystery.
EATR-flooding unlocks accurate rate constant calculations for biomolecular processes using static biasing schemes, overcoming a key limitation of previous EATR methods.
GNNs can predict core-electron binding energies in organic molecules with surprising accuracy (0.33 eV error), offering a computationally efficient alternative to expensive quantum chemistry calculations.
Radicals are the unsung heroes of soot formation, orchestrating the transition from disordered ring-edge structures to thermally stable, electron-delocalized architectures.
Transfer learning can unlock scalable emission control across diverse waste incineration plants by learning transferable system-level structures that capture physical constraints, operating-regime heterogeneity, and carbon-pollutant coupling.
Silicon anodes don't just degrade faster, they degrade *differently*, and this physics-based model reveals how.
Unlocking the 10-12.5 μm molecular fingerprint region for microsecond-resolved spectroscopy opens new avenues for studying the kinetics of short-lived radicals.
Occupancy extrapolation can now compete with Bethe-Salpeter for excited state calculations, but with better performance on triplet and Rydberg states.
Traditional research papers are costing AI agents reproducibility and understanding, but a new "Agent-Native" format that captures the full messy research process boosts performance by up to 20%.
Ditch softmax attention for sigmoid: it unlocks 25% better cell-type separation, 10% faster training, and rock-solid stability for biological foundation models.
Unlock chemical accuracy for strongly correlated systems without sacrificing computational cost: SZ-QCT extends canonical transformation theory to include four-body interactions, achieving sub-millihartree errors with $\mathcal{O}(N^8/n_c)$ scaling.
Forget exponentially scaling complexity: representing realistic atomic orbitals on quantum computers might be easier than we thought, thanks to bounded entanglement in Matrix Product State encodings.
Photocatalysis can bypass kinetic bottlenecks in styrene production, but only if you tune the wavelength to match the excited state landscape of the TiO2 catalyst.
Depression leaves a detectable fingerprint in the way our vocal system revisits acoustic states during conversation, revealing new avenues for digital biomarkers.
An AI agent autonomously discovered four new superconductors, shrinking the discovery timeline from years to GPU hours.
Catch AI's academic dishonesty: HalluCiteChecker spots bogus citations in seconds, lightening the load for reviewers drowning in AI-assisted papers.
LLMs that ace general web browsing still fail miserably at autonomous scientific literature discovery, revealing a critical gap in research-oriented AI agent capabilities.
GNNs tagging jets at the LHC aren't black boxes: explainability methods reveal they learn physically meaningful features of QCD, with performance varying predictably across energy regimes.
Self-attention can emerge naturally from the competitive dynamics of neuron-astrocyte interactions, offering a biologically plausible alternative to standard attention mechanisms.
Differentiable physics unlocks adaptable and scalable phase retrieval for coherent transition radiation spectroscopy, outperforming traditional methods by seamlessly incorporating complex experimental effects.
Skip the retraining: AM-SGHMC lets you apply a single trained MCMC sampler to various Bayesian updating problems for similar structures.
Stop hand-tuning kernels for Koopman operator approximation: this dictionary learning approach automatically discovers optimal kernel parameters, simplifying kEDMD.
PINNs can now jointly infer parameters and detect regime-switching in dynamical systems, outperforming traditional methods that treat these tasks separately.
Learning quantum dynamics through Bohmian trajectories transforms the time-dependent Schrödinger equation into a self-consistent generative modeling problem.
Generate calibrated, high-resolution precipitation ensembles from coarse climate models on your laptop in seconds, outperforming traditional downscaling methods by a wide margin.
COBALT unlocks efficient structural design optimization by treating the design space as a discrete anchored graph, avoiding the pitfalls of continuous relaxation and rounding-off that plague existing methods.
Forget struggling with non-convex optimization for causal discovery: a new algorithm extracts causal order directly from score functions with simple matrix operations.
Accurate landslide prediction is possible with sparse data by injecting geomorphic priors, unlocking geohazard risk assessment in data-scarce mountainous regions.
Achieve superior PDE solution accuracy with Shearlet Neural Operators, which leverage directional multiscale analysis to capture anisotropic shocks and discontinuities that hamstring Fourier-based methods.
Generating realistic landslide datasets from sparse, imbalanced real-world data is now possible, thanks to a tabular foundation model that captures complex feature dependencies.
Kohn-Sham eigenvalues, often dismissed as unphysical, actually represent quasiparticle bands after accounting for dynamical core excitations, resolving a decades-old discrepancy with ARPES measurements.
Popular dimensionality reduction techniques like UMAP can *invent* topological structure not present in the original data, but DiRe avoids this pitfall while matching UMAP's speed and classification performance.
LLMs crush traditional methods in recipe nutrient estimation, but the accuracy boost comes at a steep computational price, forcing a trade-off between precision and real-time performance.
Automating RCT benchmarking and observational trial calibration could unlock more reliable real-world evidence for clinical decision-making.
Current machine learning models for semiconductor bandgap prediction fall short when faced with the messy reality of experimental data, highlighting a critical need for more robust and generalizable learning strategies.
LLMs can now automatically design and execute experiments to resolve debates between cognitive science theories, even discovering the models and experiments themselves.
Dutch NLP researchers, rejoice: a massive, freely available 35B token medical corpus has arrived to jumpstart your models.
Agentic AI systems can confidently generate plausible but wrong scientific results, even when given domain-specific context, highlighting a critical challenge for their integration into research workflows.
Fragmented medical data hurts MLLM performance: this paper shows how a hierarchical medical knowledge graph can be used to engineer training data that substantially improves MLLM accuracy on complex clinical tasks.
LLMs struggle with clinical trial reasoning due to implicit planning assumptions, but a multi-LLM planner that explicitly decomposes the task into structured steps significantly improves accuracy and efficiency.
Non-contact pain detection in neonates is now possible using facial video analysis, opening the door to continuous, less intrusive monitoring in neonatal ICUs.
