Scientific Discovery & Drug Design - Weekly Roundup

LLM agents significantly outperform numerical optimizers in four out of five black-box optimization domains, but adding more numerical tools doesn't always help.

A single neural density functional accurately predicts ion correlations in polymer electrolytes across varying concentrations and geometries, generalizing beyond training data and outperforming existing architectures.

Reverse-mode gradients slash Bayesian inference time for expensive simulators, especially when parameters vastly outnumber outputs.

Spatio-temporal Bayesian optimization for point-process data is now scalable and efficient, achieving accurate peak discovery with minimal evaluations.

State-preserving constraints in Bayesian optimization are transformed into an efficient, anisotropic ellipsoid in design space, enabling robust optimization of complex engineering systems.

LLM agents significantly outperform numerical optimizers in four out of five black-box optimization domains, but adding more numerical tools doesn't always help.

A single neural density functional accurately predicts ion correlations in polymer electrolytes across varying concentrations and geometries, generalizing beyond training data and outperforming existing architectures.

Reverse-mode gradients slash Bayesian inference time for expensive simulators, especially when parameters vastly outnumber outputs.

Spatio-temporal Bayesian optimization for point-process data is now scalable and efficient, achieving accurate peak discovery with minimal evaluations.

State-preserving constraints in Bayesian optimization are transformed into an efficient, anisotropic ellipsoid in design space, enabling robust optimization of complex engineering systems.

A training-free change detection method achieves near-optimal performance by integrating diffusion models to precisely estimate Hyvärinen scores, slashing detection delays by over 90% in simulations.

A score-based method accelerates cluster DAG learning from interventions by two orders of magnitude, matching state-of-the-art edge recovery even on dense graphs.

Unifying factual support, counterfactual sensitivity, and exemplar tolerance reveals how GNNs truly reason about molecular properties, not just predict them.

Learnable shape parameterizations unlock superior sample efficiency and design exploration in Bayesian optimization, outperforming hand-crafted approaches across diverse engineering domains.

Research reports can now be continuously updated with over 60% fewer search calls and 33% less token consumption, maintaining high quality over time.

LLMs adapt neural operator configurations using both a strong task-dependent prior and real-time experimental feedback, outperforming Bayesian optimization.

Generative models can now capture the full spectrum of solutions in high-dimensional physical systems with symmetry-breaking bifurcations, a feat previously intractable.

Kernel methods finally offer a scalable and interpretable alternative to black-box ANNs for learning Koopman embeddings, matching or surpassing ANN performance on complex dynamical systems.

Large observational studies can finally boost the precision of small RCTs for treatment effect estimation without sacrificing unbiasedness or making strong confounding assumptions.

Auxiliary "ghost tasks" make any linear differential equation system effectively parametrizable for Gaussian Processes, solving a key limitation in physics-informed ML.

LLMs can be steered to reliably interpret complex genomic variants by grounding their reasoning in deterministic biological computations and evidence verification, avoiding hallucinated TF binding changes and fabricated experimental support.

Chronologically consistent language models trained on scientific literature and patents outperform general domain models and even predecessors on temporal science and technology tasks, even with limited historical data.

Spectral learning unlocks $\mathcal{O}(d\log d)$ covariance estimation for dense Gaussian dependencies, outperforming prior methods in forecasting and prediction tasks.

Multiplicative composition of Fourier-Chebyshev features unlocks richer PDE solution representations than implicit MLP composition.

AI agents are now building climate models that outperform published research and offer novel hypotheses for long-standing scientific debates.

Pretraining neural CFD surrogates across diverse domains unlocks substantial gains in zero- and few-shot generalization, drastically reducing data requirements for new applications.

Generative AI now predicts complex 3D urban wind and temperature fields with LES-level accuracy in seconds, unlocking rapid, turbulence-aware urban design.

A unified, open-source framework and a massive 3.3 TB dataset now enable rigorous ML-driven prediction of complex multiphase flows in porous media, accelerating research in critical energy and engineering applications.

Physics-aligned electronic ground-state models shatter out-of-distribution generalization barriers for MLIPs, achieving near-chemical accuracy on reactive systems with minimal data.

Model selection for scientific forecasting is fundamentally protocol-dependent, with optimal choices reversing based on dataset size, system complexity, and evaluation metrics like state refresh versus closed-loop performance.

Machine learning agents discover new exact values for notoriously difficult-to-compute link invariants like slice genus and unknotting number.

Generative orthogonal learning unlocks rate double robustness and quasi-oracle efficiency for estimating distributional treatment effects under complex time-varying confounding.

Bayesian inference for complex switching stochastic differential equations is now broadly tractable, even with non-analytic transition densities.

A single information-theoretic framework elegantly unifies multi-objective, multi-fidelity, and constrained Bayesian optimization, bypassing separate heuristics for each.

The asymptotic behavior of RBF kernel MLEs is finally understood, revealing minimax optimal convergence rates for all key parameters.

AI is poised to revolutionize ARPES, moving beyond data analysis to enable fully autonomous, self-driving experimental labs.

Kinetic Langevin diffusions slash posterior sampling time for imaging inverse problems by 10x while maintaining reconstruction quality, even with complex diffusion model priors.

Generative models can now produce inorganic crystals with oxidation-state distributions mirroring real-world synthesis, significantly boosting discovery potential.

A two-input Fredholm backstepping design achieves rapid stabilization of the Kuramoto--Sivashinsky equation, even with spatially varying anti-diffusion and unstable double eigenvalues, by rendering the plant controllable via one input before stabilizing with the other.

Misspecifying the coupling mechanism for ordinal data in GP-SCMs inflates counterfactual error threefold, even when observational fit is high, and this error plateaus with more data.

Complex-valued neural networks, powered by Wirtinger autodifferentiation and FFT-Hadamard convolutions, match or exceed real-valued FNet performance on language modeling with half the training steps.

Transcriptomics-specific pretraining, not model scale, unlocks state-of-the-art clinical outcome prediction from bulk RNA-seq, even for survival tasks, with a 4.2M parameter model.

Rare event probabilities in diffusion models can now be estimated orders of magnitude faster, unlocking their use in high-stakes scientific simulation.

Meta-learning unlocks efficient, resolution-independent PDE surrogate modeling for time-dependent systems by treating initial conditions as context, structuring latent spaces, and accelerating inference.

Encoding sensor geometry as a flow-matching prior dramatically improves generative models for brain signals, outperforming isotropic noise and even empirical data covariance.

AI rapidly identified differentiated peptide depot formulations with tunable 30-day release profiles by navigating a complex multidimensional design space.

Connectome-based GNNs achieve high accuracy but are dangerously overconfident, with misclassified patients receiving up to 95% confidence.

Unlocking broad downstream generalization for 3D atomistic systems, Atom-JEPA achieves state-of-the-art property prediction by learning from unlabeled structures alone.

Pinpoint the exact birth-death plane regions driving topological data differences with guaranteed error control, moving beyond global statistical tests.

Generative transport reveals a hidden time-reflection symmetry in interpolants, enabling exact, non-parametric two-sample testing without density estimation.

Generate valid and unique discrete designs in a single forward pass with over $50\times$ the yield per operation compared to prior one-step methods.

LLMs can now generate novel research ideas by explicitly learning to synthesize literature through structured "anchor" specifications, outperforming prior prompting and feedback methods.

Amortized Bayesian causal inference for hidden confounding is now possible via a single transformer pass, achieving state-of-the-art CATE estimation without hyperparameter tuning.

Chemical reaction models can now strictly enforce valence conservation and enabling rules without any training, achieving superior accuracy on reaction prediction tasks.

Discovering molecules with rare properties can now be an order of magnitude faster by testing complex mixtures and deconvoluting activity, slashing experimental costs from $\mathcal{O}(d)$ to $\mathcal{O}(\log d)$ or $\mathcal{O}(1)$.

Frozen medical AI models can now continuously learn and improve from real-world deployment experience without requiring weight updates or risking catastrophic forgetting.

AI agents in a simulated scientific ecosystem rediscovered over 60% of key findings from recent research papers, suggesting a viable path to autonomous open-ended scientific discovery.

Automated code generation from 3DGS papers not

Generating molecular graphs in a pretrained VAE's latent space via flow matching unlocks significant efficiency gains and enables property-guided generation without retraining.

LLMs can now autonomously debug complex chip designs by translating natural language queries into SQL against massive simulation dumps.

Wrist accelerometers can now yield population-scale sleep heart rate insights, with a novel transformer model outperforming existing methods in estimation availability.

Neural PDE solvers can miss optimal solutions due to unbounded parameters, but adding kernel derivatives recovers these "missing limits."

Learned AI simulators can pass standard forecast metrics yet fundamentally fail to capture system dynamics, as demonstrated by a new response theory probe that reveals critical divergences in forced responses.

Bosonic Gaussian states can be identified with unprecedented accuracy using a novel warm-start strategy combined with non-Gaussian measurements, outperforming classical Gaussian estimation.

A novel two-stream architecture disentangles temporal context from spatial refinement in EEG source imaging, achieving superior localization accuracy without compromising inference speed.

Even with only uniform state preparation and measurements, or computational basis access, we can fully reconstruct generic 2-local and many common nearest-neighbor Hamiltonians.

Explicitly linking multimodal patient data to biomedical knowledge graphs unlocks significant gains for clinical LLMs, with performance plummeting when these crucial connections are removed.

Single-neuron models can now capture multimodal voltage responses, including coexisting firing/non-firing states and depolarization block modes, a feat previously requiring massive ensembles or failing with neural operators.

Deterministic weather foundation models can be efficiently transformed into powerful probabilistic forecasters, generating skillful 15-day forecasts in minutes.

Nighttime anomaly detection is significantly enhanced by fusing M-band and DNB satellite signals, achieving higher accuracy and broader applicability than thermal-only methods.

Transformer attention applied to geometry-aware segment tokens unlocks 2459x speedup for electromigration stress prediction with sub-8% error.

Multi-omics cancer analysis becomes practical even with missing data, achieving state-of-the-art reconstruction and classification via a novel representation learning approach.

Pretrained time-series foundation models offer surprisingly inconsistent benefits for EMG tasks, with full fine-tuning often outperforming linear probing and few-shot adaptation only partially recovering performance on new users.

Equivariant potentials can achieve transformer-level accuracy with 35x fewer parameters and 10x faster inference by leveraging Cartesian tensor products.

Kernel ensembles with Survival Forests and Beran estimators significantly outperform single models for patient-specific survival prediction.

Personalized suicide risk forecasting is significantly improved by embedding patients in a latent space that models similarity, outperforming pooled and per-patient models by up to 30% for first-occurrence event prediction.

Deterministic midpoint methods for Langevin sampling can achieve superior sample complexity in high dimensions compared to their randomized counterparts.

Unreported sinkhole locations are effectively treated as right-censored, unlocking continuous, time-aware susceptibility predictions from multimodal data.

Closed-form KANs integrated into IGA dramatically slash PDE solution errors by orders of magnitude, outperforming pure neural solvers and standard IGA.

Unsupervised EEG domain adaptation is unlocked by treating distribution shifts as linear transformations on the SPD manifold, enabling interpretable alignment without labeled calibration data.

Self-supervised learning on mass spectrometry data unlocks state-of-the-art de novo peptide sequencing, outperforming fully supervised methods without any sequence labels.

Frontier AI models now achieve human-level experimental research taste, outperforming experts with 30x less cost and demonstrating a doubling of compute efficiency every 3 months.

Language models can autonomously discover superior optimization algorithms for scientific computing, outperforming state-of-the-art methods by up to 59.8% in force-call efficiency for molecular geometry optimization.

NeuroCBIR achieves near-perfect subject re-identification in 3D brain MRIs and encodes clinically relevant information for downstream tasks, all within seconds per scan.