Search

Correlative computational microscopy is accelerating the mapping of dynamic biological systems by integrating morphological and molecular measurements across spatial scales, from organelles to entire organisms. Visualization, measurement, and prediction of interactions among the components of biological systems can be accelerated by generalist computational imaging frameworks that relax the trade-offs imposed by multiplex dynamic imaging. This work reports a generalist framework for wave optical imaging of the architectural order (waveOrder) among biomolecules for encoding and decoding multiple specimen properties from a minimal set of acquired channels, with or without fluorescent labels. waveOrder expresses material properties in terms of elegant physically motivated basis vectors directly interpretable as phase, absorption, birefringence, diattenuation, and fluorophore density; and it expresses image data in terms of directly measurable Stokes parameters. We report a corresponding multi-channel reconstruction algorithm to recover specimen properties in multiple contrast modes. With this framework, we implement multiple 3D computational microscopy methods, including quantitative phase imaging, quantitative label-free imaging with phase and polarization, and fluorescence deconvolution imaging, across scales ranging from organelles to whole zebrafish. These advances are available via an extensible open-source computational imaging library, waveOrder, and a napari plugin, recOrder., Comment: 11 pages of main text (5 figures, one table); 9 pages of supplementary text (4 figures, one table); 5 ancillary videos

Total variation (TV) is a widely used function for regularizing imaging inverse problems that is particularly appropriate for images whose underlying structure is piecewise constant. TV regularized optimization problems are typically solved using proximal methods, but the way in which they are applied is constrained by the absence of a closed-form expression for the proximal operator of the TV function. A closed-form approximation of the TV proximal operator has previously been proposed, but its accuracy was not theoretically explored in detail. We address this gap by making several new theoretical contributions, proving that the approximation leads to a proximal operator of some convex function, that it always decreases the TV function, and that its error can be fully characterized and controlled with its scaling parameter. We experimentally validate our theoretical results on image denoising and sparse-view computed tomography (CT) image reconstruction.

Causal discovery aims to automatically uncover causal relationships from data, a capability with significant potential across many scientific disciplines. However, its real-world applications remain limited. Current methods often rely on unrealistic assumptions and are evaluated only on simple synthetic toy datasets, often with inadequate evaluation metrics. In this paper, we substantiate these claims by performing a systematic review of the recent causal discovery literature. We present applications in biology, neuroscience, and Earth sciences - fields where causal discovery holds promise for addressing key challenges. We highlight available simulated and real-world datasets from these domains and discuss common assumption violations that have spurred the development of new methods. Our goal is to encourage the community to adopt better evaluation practices by utilizing realistic datasets and more adequate metrics., Comment: 39 pages, 8 figures

Enabling effective collaboration among LLMs is a crucial step toward developing autonomous systems capable of solving complex problems. While LLMs are typically used as single-model generators, where humans critique and refine their outputs, the potential for jointly-trained collaborative models remains largely unexplored. Despite promising results in multi-agent communication and debate settings, little progress has been made in training models to work together on tasks. In this paper, we present a first step toward "Multi-agent LLM training" (MALT) on reasoning problems. Our approach employs a sequential multi-agent setup with heterogeneous LLMs assigned specialized roles: a generator, verifier, and refinement model iteratively solving problems. We propose a trajectory-expansion-based synthetic data generation process and a credit assignment strategy driven by joint outcome based rewards. This enables our post-training setup to utilize both positive and negative trajectories to autonomously improve each model's specialized capabilities as part of a joint sequential system. We evaluate our approach across MATH, GSM8k, and CQA, where MALT on Llama 3.1 8B models achieves relative improvements of 14.14%, 7.12%, and 9.40% respectively over the same baseline model. This demonstrates an early advance in multi-agent cooperative capabilities for performance on mathematical and common sense reasoning questions. More generally, our work provides a concrete direction for research around multi-agent LLM training approaches., Comment: Preliminary work

The Vera C. Rubin Observatory is due to commence the 10-year Legacy Survey of Space and Time (LSST) at the end of 2025. To detect transient/variable sources and identify solar system objects (SSOs), the processing pipelines require templates of the static sky to perform difference imaging. During the first year of the LSST, templates must be generated as the survey progresses, otherwise SSOs cannot be discovered nightly. The incremental template generation strategy has not been finalized; therefore, we use the Metric Analysis Framework (MAF) and a simulation of the survey cadence (one_snap_v4.0_10yrs}) to explore template generation in Year 1. We have assessed the effects of generating templates over timescales of days-weeks, when at least four images of sufficient quality are available for $\geq90\%$ of the visit. We predict that SSO discoveries will begin $\sim$2-3 months after the start of the survey. We find that the ability of the LSST to discover SSOs in real-time is reduced in Year 1. This is especially true for detections in areas of the sky that receive fewer visits, such as the North Ecliptic Spur (NES), and in less commonly used filters, such as the $u$ and $g$-bands. The lack of templates in the NES dominates the loss of real-time SSO discoveries; across the whole sky the MAF Main-Belt asteroid (MBA) discovery metric decreases by up to $63\%$ compared to the baseline observing strategy, whereas the metric decreases by up to $79\%$ for MBAs in the NES alone., Comment: Submitted for consideration in the ApJS Rubin LSST Survey Strategy Optimization Focus Issue

Motivated by the scarcity of proper labels in an astrophysical application, we have developed a novel technique, called Selfish Evolution, which allows for the detection and correction of corrupted labels in a weakly supervised fashion. Unlike methods based on early stopping, we let the model train on the noisy dataset. Only then do we intervene and allow the model to overfit to individual samples. The ``evolution'' of the model during this process reveals patterns with enough information about the noisiness of the label, as well as its correct version. We train a secondary network on these spatiotemporal ``evolution cubes'' to correct potentially corrupted labels. We incorporate the technique in a closed-loop fashion, allowing for automatic convergence towards a mostly clean dataset, without presumptions about the state of the network in which we intervene. We evaluate on the main task of the Supernova-hunting dataset but also demonstrate efficiency on the more standard MNIST dataset.

AI models are increasingly prevalent in high-stakes environments, necessitating thorough assessment of their capabilities and risks. Benchmarks are popular for measuring these attributes and for comparing model performance, tracking progress, and identifying weaknesses in foundation and non-foundation models. They can inform model selection for downstream tasks and influence policy initiatives. However, not all benchmarks are the same: their quality depends on their design and usability. In this paper, we develop an assessment framework considering 46 best practices across an AI benchmark's lifecycle and evaluate 24 AI benchmarks against it. We find that there exist large quality differences and that commonly used benchmarks suffer from significant issues. We further find that most benchmarks do not report statistical significance of their results nor allow for their results to be easily replicated. To support benchmark developers in aligning with best practices, we provide a checklist for minimum quality assurance based on our assessment. We also develop a living repository of benchmark assessments to support benchmark comparability, accessible at betterbench.stanford.edu., Comment: Accepted as a Spotlight Poster to NeurIPS 2024

We present a hierarchical framework to solve robot planning as an input control problem. At the lowest level are temporary closed control loops, ("tasks"), each representing a behaviour, contingent on a specific sensory input and therefore temporary. At the highest level, a supervising "Configurator" directs task creation and termination. Here resides "core" knowledge as a physics engine, where sequences of tasks can be simulated. The Configurator encodes and interprets simulation results,based on which it can choose a sequence of tasks as a plan. We implement this framework on a real robot and test it in an overtaking scenario as proof-of-concept.

Bending the edge of a thin elastic material promotes rigidity far from its clamped boundary. However, this curvature-induced rigidity can be overwhelmed by gravity or other external loading, resulting in elastic buckling and large deformations. We consider the role of body geometry on this competition using experiments, numerical simulations, and reduced-order models. Finite element simulations are performed using a model nonlinear hyperelastic material, and a theoretical framework is proposed that incorporates small lateral curvatures, large longitudinal rotations, and a varying cross-sectional width. A particular focus is on the comparison between rectangular and triangular sheets, and trapezoidal sheets in between. Sheet geometry affects downward tip deflection by changing the relative importance of the sheet's weight and the rigidity provided by curvature, often in subtle ways. In extreme cases, non-monotonic deflection is observed with increasing sheet length, and a region of hysteretic bistability emerges, becoming more pronounced with rectangular sheets and large imposed curvatures. These findings demonstrate the profound impact of geometry on the competition between curvature-induced rigidity and gravity-induced deformation in thin elastic materials., Comment: 17 pages, 10 figures

Accurate estimates of causal effects play a key role in decision-making across applications such as healthcare, economics, and operations. In the absence of randomized experiments, a common approach to estimating causal effects uses \textit{covariate adjustment}. In this paper, we study covariate adjustment for discrete distributions from the PAC learning perspective, assuming knowledge of a valid adjustment set $\bZ$, which might be high-dimensional. Our first main result PAC-bounds the estimation error of covariate adjustment by a term that is exponential in the size of the adjustment set; it is known that such a dependency is unavoidable even if one only aims to minimize the mean squared error. Motivated by this result, we introduce the notion of an \emph{$\eps$-Markov blanket}, give bounds on the misspecification error of using such a set for covariate adjustment, and provide an algorithm for $\eps$-Markov blanket discovery; our second main result upper bounds the sample complexity of this algorithm. Furthermore, we provide a misspecification error bound and a constraint-based algorithm that allow us to go beyond $\eps$-Markov blankets to even smaller adjustment sets. Our third main result upper bounds the sample complexity of this algorithm, and our final result combines the first three into an overall PAC bound. Altogether, our results highlight that one does not need to perfectly recover causal structure in order to ensure accurate estimates of causal effects.

We report the detection of water vapor associated with main-belt comet 358P/PANSTARRS on UT 2024 January 8-9 using the NIRSPEC instrument aboard JWST. We derive a water production rate of Q(H2O)=(5.0+/-0.2)x10^25 molecules/s, marking only the second direct detection of sublimation products of any kind from a main-belt comet, after 238P/Read. Similar to 238P, we find a remarkable absence of hypervolatile species, finding Q(CO2)<7.6x10^22 molecules/s, corresponding to Q(CO2)/Q(H2O)<0.2%. Upper limits on CH3OH and CO emission are also estimated. Photometry from ground-based observations show that the dust coma brightened and faded slowly over ~250 days in 2023-2024, consistent with photometric behavior observed in 2012-2013, but also indicate a ~2.5x decline in the dust production rate between these two periods. Dynamical dust modeling shows that the coma's morphology as imaged by JWST's NIRCAM instrument on 2023 November 22 can be reproduced by asymmetric dust emission from a nucleus with a mid-range obliquity (~80 deg) with a steady-state mass loss rate of ~0.8 kg/s. Finally, we find similar Afrho-to-gas ratios of log10(Afrho/Q(H2O))=-24.8+/-0.2 for 358P and log10(Afrho/QH2O)=-24.4+/-0.2 for 238P, suggesting that Afrho could serve as an effective proxy for estimating water production rates in other active main-belt comets. The confirmation of water vapor outgassing in both main-belt comets observed by JWST to date reinforces the use of recurrent activity near perihelion as an indicator of sublimation-driven activity in active asteroids., Comment: 32 pages, 13 figures. Accepted for publication in The Planetary Science Journal

The problem of determining the configuration of points from partial distance information, known as the Euclidean Distance Geometry (EDG) problem, is fundamental to many tasks in the applied sciences. In this paper, we propose two algorithms grounded in the Riemannian optimization framework to address the EDG problem. Our approach formulates the problem as a low-rank matrix completion task over the Gram matrix, using partial measurements represented as expansion coefficients of the Gram matrix in a non-orthogonal basis. For the first algorithm, under a uniform sampling with replacement model for the observed distance entries, we demonstrate that, with high probability, a Riemannian gradient-like algorithm on the manifold of rank-$r$ matrices converges linearly to the true solution, given initialization via a one-step hard thresholding. This holds provided the number of samples, $m$, satisfies $m \geq \mathcal{O}(n^{7/4}r^2 \log(n))$. With a more refined initialization, achieved through resampled Riemannian gradient-like descent, we further improve this bound to $m \geq \mathcal{O}(nr^2 \log(n))$. Our analysis for the first algorithm leverages a non-self-adjoint operator and depends on deriving eigenvalue bounds for an inner product matrix of restricted basis matrices, leveraging sparsity properties for tighter guarantees than previously established. The second algorithm introduces a self-adjoint surrogate for the sampling operator. This algorithm demonstrates strong numerical performance on both synthetic and real data. Furthermore, we show that optimizing over manifolds of higher-than-rank-$r$ matrices yields superior numerical results, consistent with recent literature on overparameterization in the EDG problem., Comment: 38 pages, 4 figures, 5 tables

Language models (LMs) can exhibit human-like behaviour, but it is unclear how to describe this behaviour without undue anthropomorphism. We formalise a behaviourist view of LM character traits: qualities such as truthfulness, sycophancy, or coherent beliefs and intentions, which may manifest as consistent patterns of behaviour. Our theory is grounded in empirical demonstrations of LMs exhibiting different character traits, such as accurate and logically coherent beliefs, and helpful and harmless intentions. We find that the consistency with which LMs exhibit certain character traits varies with model size, fine-tuning, and prompting. In addition to characterising LM character traits, we evaluate how these traits develop over the course of an interaction. We find that traits such as truthfulness and harmfulness can be stationary, i.e., consistent over an interaction, in certain contexts, but may be reflective in different contexts, meaning they mirror the LM's behavior in the preceding interaction. Our formalism enables us to describe LM behaviour precisely in intuitive language, without undue anthropomorphism., Comment: accepted as Findings of EMNLP2024

Superfast rotators (SFRs) are small solar system objects that rotate faster than generally possible for a cohesionless rubble pile. Their rotational characteristics allow us to make inferences about their interior structure and composition. Here, we present the methods and results from a preliminary search for SFRs in the DECam Ecliptic Exploration Project (DEEP) data set. We find three SFRs from a sample of 686 main-belt asteroids, implying an occurrence rate of 0.4 -0.3/+0.1 percent - a higher incidence rate than has been measured by previous studies. We suggest that this high occurrence rate is due to the small sub-kilometer size regime to which DEEP has access: the objects searched here were as small as 500 m. We compute the minimum required cohesive strength for each of these SFRs and discuss the implications of these strengths in the context of likely evolution mechanisms. We find that all three of these SFRs require strengths that are more than that of weak regolith but consistent with many cohesive asteroid strengths reported in the literature. Across the full DEEP data set, we have identified ~70,000 Main-Belt Asteroids and expect ~300 SFRs - a result that will be assessed in a future paper.

A body immersed in a nematic liquid crystal disturbs the fluid's preferred molecular configuration and increases its stored elastic energy. In an active nematic, the fluid components also generate a stress in the bulk fluid. By introducing either an immersed body or boundary, a fluid flow can be triggered due to anchoring boundary conditions. The fluid imposes elastic, viscous, and active stresses on such surfaces, which, if compliant, may result in a surface deformation. We study the deformations and stress in a linearly elastic body placed in an active nematic in two-dimensions. Using complex variables techniques, exact expressions for the fluid flow, director field, surface tractions, and body deformation are derived. Qualitative differences between elastic and active stress-driven deformations are identified, depending on an active Ericksen number, anchoring conditions, and body material properties, thereby suggesting a new method for measuring the active stresses in active anisotropic biological systems. Flow profiles, external confinement, and anchoring-induced stirring are also addressed., Comment: Paper: 8 pages, 4 figures; Supplemental Material: 7 pages

Everyday experience shows that cellular sheets are stiffened by the presence of a pressurized gas: from bicycle inner tubes to bubble wrap, the presence of an internal pressure increases the stiffness of otherwise floppy structures. The same is true of plants, with turgor pressure (due to the presence of water) taking the place of gas pressure; indeed, in the absence of water, many plants wilt. However, the mechanical basis of this stiffening is somewhat opaque: simple attempts to rationalize it suggest that the stiffness should be independent of the pressure, at odds with everyday experience. Here, we study the mechanics of sheets that are a single cell thick and show how a pressure-dependent bending stiffness may arise. Our model rationalizes observations of turgor-driven shrinkage in plant cells and also suggests that turgor is unlikely to provide significant structural support in many monolayer leaves, such as those found in mosses. However, for such systems, turgor does provide a way to control leaf shape, in accordance with observations of curling upon drying of moss leaves. Guided by our results, we also present a biomimetic actuator that un-curls upon pressurization., Comment: Paper: 14 pages, 9 figures; Supplemental Material: 8 pages, 2 figures

Dynamic maneuvers for legged robots present a difficult challenge due to the complex dynamics and contact constraints. This paper introduces a versatile trajectory optimization framework for continuous-time multi-phase problems. We introduce a new transcription scheme that enables pseudospectral collocation to optimize directly on Lie Groups, such as SE(3) and quaternions without special normalization constraints. The key insight is the change of variables - we choose to optimize over the history of the tangent vectors rather than the states themselves. Our approach uses a modified Legendre-Gauss-Radau (LGR) method to produce dynamic motions for various legged robots. We implement our approach as a Model Predictive Controller (MPC) and track the MPC output using a Quadratic Program (QP) based whole-body controller. Results on the Go1 Unitree and WPI HURON humanoid confirm the feasibility of the planned trajectories., Comment: This extended abstract was accepted for presentation at ICRA@40

This paper explores the concept of external magnetic control for vine robots to enable their high curvature steering and navigation for use in endoluminal applications. Vine robots, inspired by natural growth and locomotion strategies, present unique shape adaptation capabilities that allow passive deformation around obstacles. However, without additional steering mechanisms, they lack the ability to actively select the desired direction of growth. The principles of magnetically steered growing robots are discussed, and experimental results showcase the effectiveness of the proposed magnetic actuation approach. We present a 25 mm diameter vine robot with integrated magnetic tip capsule, including 6 Degrees of Freedom (DOF) localization and camera and demonstrate a minimum bending radius of 3.85 cm with an internal pressure of 30 kPa. Furthermore, we evaluate the robot's ability to form tight curvature through complex navigation tasks, with magnetic actuation allowing for extended free-space navigation without buckling. The suspension of the magnetic tip was also validated using the 6 DOF localization system to ensure that the shear-free nature of vine robots was preserved. Additionally, by exploiting the magnetic wrench at the tip, we showcase preliminary results of vine retraction. The findings contribute to the development of controllable vine robots for endoluminal applications, providing high tip force and shear-free navigation., Comment: 13 pages, 10 figures

Molecular deuteration is a powerful diagnostic tool for probing the physical conditions and chemical processes in astrophysical environments. In this work, we focus on formaldehyde deuteration in the protobinary system NGC\,1333 IRAS\,4A, located in the Perseus molecular cloud. Using high-resolution ($\sim$\,100\,au) ALMA observations, we investigate the [D$_2$CO]/[HDCO] ratio along the cavity walls of the outflows emanating from IRAS\,4A1. Our analysis reveals a consistent decrease in the deuteration ratio (from $\sim$\,60-20\% to $\sim$\,10\%) with increasing distance from the protostar (from $\sim$\,2000\,au to $\sim$\,4000\,au). Given the large measured [D$_2$CO]/[HDCO], both HDCO and D$_2$CO are likely injected by the shocks along the cavity walls into the gas-phase from the dust mantles, formed in the previous prestellar phase. We propose that the observed [D$_2$CO]/[HDCO] decrease is due to the density profile of the prestellar core from which NGC\,1333 IRAS\,4A was born. When considering the chemical processes at the base of formaldehyde deuteration, the IRAS\,4A's prestellar precursor had a predominantly flat density profile within 3000\,au and a decrease of density beyond this radius.