Search

Recent advances in aligning Large Language Models with human preferences have benefited from larger reward models and better preference data. However, most of these methodologies rely on the accuracy of the reward model. The reward models used in Reinforcement Learning with Human Feedback (RLHF) are typically learned from small datasets using stochastic optimization algorithms, making them prone to high variability. We illustrate the inconsistencies between reward models empirically on numerous open-source datasets. We theoretically show that the fluctuation of the reward models can be detrimental to the alignment problem because the derived policies are more overfitted to the reward model and, hence, are riskier if the reward model itself is uncertain. We use concentration of measure to motivate an uncertainty-aware, conservative algorithm for policy optimization. We show that such policies are more risk-averse in the sense that they are more cautious of uncertain rewards. We theoretically prove that our proposed methodology has less risk than the vanilla method. We corroborate our theoretical results with experiments based on designing an ensemble of reward models. We use this ensemble of reward models to align a language model using our methodology and observe that our empirical findings match our theoretical predictions.

Early screening for cancer has proven to improve the survival rate and spare patients from intensive and costly treatments due to late diagnosis. Cancer screening in the healthy population involves an initial risk stratification step to determine the screening method and frequency, primarily to optimize resource allocation by targeting screening towards individuals who draw most benefit. For most screening programs, age and clinical risk factors such as family history are part of the initial risk stratification algorithm. In this paper, we focus on developing a blood marker-based risk stratification approach, which could be used to identify patients with elevated cancer risk to be encouraged for taking a diagnostic test or participate in a screening program. We demonstrate that the combination of simple, widely available blood tests, such as complete blood count and complete metabolic panel, could potentially be used to identify patients at risk for colorectal, liver, and lung cancers with areas under the ROC curve of 0.76, 0.85, 0.78, respectively. Furthermore, we hypothesize that such an approach could not only be used as pre-screening risk assessment for individuals but also as population health management tool, for example to better interrogate the cancer risk in certain sub-populations.

Humans can perform various combinations of physical skills without having to relearn skills from scratch every single time. For example, we can swing a bat when walking without having to re-learn such a policy from scratch by composing the individual skills of walking and bat swinging. Enabling robots to combine or compose skills is essential so they can learn novel skills and tasks faster with fewer real world samples. To this end, we propose a novel compositional approach called DSE- Diffusion Score Equilibrium that enables few-shot learning for novel skills by utilizing a combination of base policy priors. Our method is based on probabilistically composing diffusion policies to better model the few-shot demonstration data-distribution than any individual policy. Our goal here is to learn robot motions few-shot and not necessarily goal oriented trajectories. Unfortunately we lack a general purpose metric to evaluate the error between a skill or motion and the provided demonstrations. Hence, we propose a probabilistic measure - Maximum Mean Discrepancy on the Forward Kinematics Kernel (MMD-FK), that is task and action space agnostic. By using our few-shot learning approach DSE, we show that we are able to achieve a reduction of over 30% in MMD-FK across skills and number of demonstrations. Moreover, we show the utility of our approach through real world experiments by teaching novel trajectories to a robot in 5 demonstrations., Comment: 6(+1) pages, 6 figures

Large language models (LLMs) have demonstrated impressive performance on several tasks and are increasingly deployed in real-world applications. However, especially in high-stakes settings, it becomes vital to know when the output of an LLM may be unreliable. Depending on whether an answer is trustworthy, a system can then choose to route the question to another expert, or otherwise fall back on a safe default behavior. In this work, we study the extent to which LLMs can reliably indicate confidence in their answers, and how this notion of confidence can translate into downstream accuracy gains. We propose Self-REF, a lightweight training strategy to teach LLMs to express confidence in whether their answers are correct in a reliable manner. Self-REF introduces confidence tokens into the LLM, from which a confidence score can be extracted. Compared to conventional approaches such as verbalizing confidence and examining token probabilities, we demonstrate empirically that confidence tokens show significant improvements in downstream routing and rejection learning tasks.

Classical manipulator motion planners work across different robot embodiments. However they plan on a pre-specified static environment representation, and are not scalable to unseen dynamic environments. Neural Motion Planners (NMPs) are an appealing alternative to conventional planners as they incorporate different environmental constraints to learn motion policies directly from raw sensor observations. Contemporary state-of-the-art NMPs can successfully plan across different environments. However none of the existing NMPs generalize across robot embodiments. In this paper we propose Cross-Embodiment Motion Policy (XMoP), a neural policy for learning to plan over a distribution of manipulators. XMoP implicitly learns to satisfy kinematic constraints for a distribution of robots and $\textit{zero-shot}$ transfers the planning behavior to unseen robotic manipulators within this distribution. We achieve this generalization by formulating a whole-body control policy that is trained on planning demonstrations from over three million procedurally sampled robotic manipulators in different simulated environments. Despite being completely trained on synthetic embodiments and environments, our policy exhibits strong sim-to-real generalization across manipulators with different kinematic variations and degrees of freedom with a single set of frozen policy parameters. We evaluate XMoP on $7$ commercial manipulators and show successful cross-embodiment motion planning, achieving an average $70\%$ success rate on baseline benchmarks. Furthermore, we demonstrate our policy sim-to-real on two unseen manipulators solving novel planning problems across three real-world domains even with dynamic obstacles., Comment: Website at https://prabinrath.github.io/xmop Paper has 17 pages, 13 figures, 5 tables

Continual and interactive robot learning is a challenging problem as the robot is present with human users who expect the robot to learn novel skills to solve novel tasks perpetually with sample efficiency. In this work we present a framework for robots to query and learn visuo-motor robot skills and task relevant information via natural language dialog interactions with human users. Previous approaches either focus on improving the performance of instruction following agents, or passively learn novel skills or concepts. Instead, we used dialog combined with a language-skill grounding embedding to query or confirm skills and/or tasks requested by a user. To achieve this goal, we developed and integrated three different components for our agent. Firstly, we propose a novel visual-motor control policy ACT with Low Rank Adaptation (ACT-LoRA), which enables the existing SoTA ACT model to perform few-shot continual learning. Secondly, we develop an alignment model that projects demonstrations across skill embodiments into a shared embedding allowing us to know when to ask questions and/or demonstrations from users. Finally, we integrated an existing LLM to interact with a human user to perform grounded interactive continual skill learning to solve a task. Our ACT-LoRA model learns novel fine-tuned skills with a 100% accuracy when trained with only five demonstrations for a novel skill while still maintaining a 74.75% accuracy on pre-trained skills in the RLBench dataset where other models fall significantly short. We also performed a human-subjects study with 8 subjects to demonstrate the continual learning capabilities of our combined framework. We achieve a success rate of 75% in the task of sandwich making with the real robot learning from participant data demonstrating that robots can learn novel skills or task knowledge from dialogue with non-expert users using our approach.

This work presents the design and development of IIT Bombay Racing's Formula Student style autonomous racecar algorithm capable of running at the racing events of Formula Student-AI, held in the UK. The car employs a cutting-edge sensor suite of the compute unit NVIDIA Jetson Orin AGX, 2 ZED2i stereo cameras, 1 Velodyne Puck VLP16 LiDAR and SBG Systems Ellipse N GNSS/INS IMU. It features deep learning algorithms and control systems to navigate complex tracks and execute maneuvers without any human intervention. The design process involved extensive simulations and testing to optimize the vehicle's performance and ensure its safety. The algorithms have been tested on a small scale, in-house manufactured 4-wheeled robot and on simulation software. The results obtained for testing various algorithms in perception, simultaneous localization and mapping, path planning and controls have been detailed., Comment: 8 pages, 19 figures

We present the first application of modern Hopfield networks to the problem of portfolio optimization. We performed an extensive study based on combinatorial purged cross-validation over several datasets and compared our results to both traditional and deep-learning-based methods for portfolio selection. Compared to state-of-the-art deep-learning methods such as Long-Short Term Memory networks and Transformers, we find that the proposed approach performs on par or better, while providing faster training times and better stability. Our results show that Modern Hopfield Networks represent a promising approach to portfolio optimization, allowing for an efficient, scalable, and robust solution for asset allocation, risk management, and dynamic rebalancing., Comment: 12 pages, 4 figures

When reasoning about formal objects whose structures involve binding, it is often necessary to analyze expressions relative to a context that associates types, values, and other related attributes with variables that appear free in the expressions. We refer to such associations as binding contexts. Reasoning tasks also require properties such as the shape and uniqueness of associations concerning binding contexts to be made explicit. The Abella proof assistant, which supports a higher-order treatment of syntactic constructs, provides a simple and elegant way to describe such contexts from which their properties can be extracted. This mechanism is based at the outset on viewing binding contexts as ordered sequences of associations. However, when dealing with object systems that embody notions of linearity, it becomes necessary to treat binding contexts more generally as partitionable multisets. We show how to adapt the original Abella encoding to encompass such a generalization. The key idea in this adaptation is to base the definition of a binding context on a mapping to an underlying ordered sequence of associations. We further show that properties that hold with the ordered sequence view can be lifted to the generalized definition of binding contexts and that this lifting can, in fact, be automated. These ideas find use in the extension currently under development of the two-level logic approach of Abella to a setting where linear logic is used as the specification logic., Comment: In Proceedings LFMTP 2024, arXiv:2407.05822

Adaptive networks can sense and adjust to dynamic environments to optimize their performance. Understanding their nanoscale responses to external stimuli is essential for applications in nanodevices and neuromorphic computing. However, it is challenging to image such responses on the nanoscale with crystallographic sensitivity. Here, the evolution of nanodomain networks in (PbTiO3)n/(SrTiO3)n superlattices was directly visualized in real space as the system adapts to ultrafast repetitive optical excitations that emulate controlled neural inputs. The adaptive response allows the system to explore a wealth of metastable states that were previously inaccessible. Their reconfiguration and competition were quantitatively measured by scanning x-ray nanodiffraction as a function of the number of applied pulses, in which crystallographic characteristics were quantitatively assessed by assorted diffraction patterns using unsupervised machine-learning methods. The corresponding domain boundaries and their connectivity were drastically altered by light, holding promise for light-programmable nanocircuits in analogy to neuroplasticity. Phase-field simulations elucidate that the reconfiguration of the domain networks is a result of the interplay between photocarriers and transient lattice temperature. The demonstrated optical control scheme and the uncovered nanoscopic insights open opportunities for remote control of adaptive nanoscale domain networks.

Projecting sea-level change in various climate-change scenarios typically involves running forward simulations of the Earth's gravitational, rotational and deformational (GRD) response to ice mass change, which requires high computational cost and time. Here we build neural-network emulators of sea-level change at 27 coastal locations, due to the GRD effects associated with future Antarctic Ice Sheet mass change over the 21st century. The emulators are based on datasets produced using a numerical solver for the static sea-level equation and published ISMIP6-2100 ice-sheet model simulations referenced in the IPCC AR6 report. We show that the neural-network emulators have an accuracy that is competitive with baseline machine learning emulators. In order to quantify uncertainty, we derive well-calibrated prediction intervals for simulated sea-level change via a linear regression postprocessing technique that uses (nonlinear) machine learning model outputs, a technique that has previously been applied to numerical climate models. We also demonstrate substantial gains in computational efficiency: a feedforward neural-network emulator exhibits on the order of 100 times speedup in comparison to the numerical sea-level equation solver that is used for training.

While the recent literature has seen a surge in the study of constrained bandit problems, all existing methods for these begin by assuming the feasibility of the underlying problem. We initiate the study of testing such feasibility assumptions, and in particular address the problem in the linear bandit setting, thus characterising the costs of feasibility testing for an unknown linear program using bandit feedback. Concretely, we test if $\exists x: Ax \ge 0$ for an unknown $A \in \mathbb{R}^{m \times d}$, by playing a sequence of actions $x_t\in \mathbb{R}^d$, and observing $Ax_t + \mathrm{noise}$ in response. By identifying the hypothesis as determining the sign of the value of a minimax game, we construct a novel test based on low-regret algorithms and a nonasymptotic law of iterated logarithms. We prove that this test is reliable, and adapts to the `signal level,' $\Gamma,$ of any instance, with mean sample costs scaling as $\widetilde{O}(d^2/\Gamma^2)$. We complement this by a minimax lower bound of $\Omega(d/\Gamma^2)$ for sample costs of reliable tests, dominating prior asymptotic lower bounds by capturing the dependence on $d$, and thus elucidating a basic insight missing in the extant literature on such problems., Comment: Spotlight presentation at ICML 2024

We unravel the distinct roles each cation plays in phase evolution, stability, and properties within Mg1/5Co1/5Ni1/5Cu1/5Zn1/5O high-entropy oxide (HEO) by integrating experimental findings, thermodynamic analyses, and first-principles predictions. Our approach is through sequentially removing one cation at a time from the five-component high-entropy oxide to create five four-component derivatives. Bulk synthesis experiments indicate that Mg, Ni, and Co act as rock salt phase stabilizers whereas only Mg and Ni enthalpically enhance single-phase rock salt stability in thin film growth; synthesis conditions dictate whether Co is a rock salt phase stabilizer or destabilizer. By examining the competing phases and oxidation state preferences using pseudo-binary phase diagrams and first-principles calculations, we resolve the stability differences between bulk and thin film for all compositions. We systematically explore HEO macroscopic property sensitivity to cation selection employing both predicted and measured optical spectra. This study establishes a framework for understanding high-entropy oxide synthesizability and properties on a per-cation basis that is broadly applicable to tailoring functional property design in other high-entropy materials., Comment: Saeed S. I. Almishal and Jacob T. Sivak contributed equally to this work

A new horn-shaped electrooptic scanner is described with significantly improved scanning sensitivity over rectangular-shaped devices. In the new device, the shape of the scanner is chosen to follow the trajectory of the beam. An example design is described that exhibits a factor of two larger scanning sensitivity than a rectangular device with comparable maximum scanning angle. Beam propagation simulations and measurements on an experimental device verify the scanner performance., Comment: 3 pages, 3 figures. IEEE Photonics Technology Letters. Author Jennifer C. Fang is currently known as Jennifer Andreoli-Fang

An analytical theory is developed for predicting the nonlinear susceptibility of ionic polarization to continuous electromagnetic waves in both bulk and strained thin film ferroelectrics. Using a perturbation method for solving the nonlinear equation of motion for ionic polarization within the framework of Landau-Ginzburg-Devonshire theory, the full second-order nonlinear susceptibility tensor is derived as a function of frequency, temperature, and strain. The theory predicts the coexistence of a significantly enhanced second-order dielectric susceptibility and a relatively low dielectric loss in BaTiO3 films with a strain-stabilized monoclinic ferroelectric phase and in a strained SrTiO3 film near its temperature-driven second-order ferroelectric-to-paraelectric phase transition. This work establishes a theoretical framework for predicting and exploiting nonlinear interactions between THz waves and ferroelectric materials, and more generally, suggests exciting opportunities to strain-engineer nonlinear dynamical properties of ferroelectrics beyond the static and quasi-static limits.

Trigonal Cr5Te8, a self-intercalated van der Waals ferromagnet with an out of plane magnetic anisotropy, has long been known to crystallise in a centrosymmetric structure. Through detailed structural analysis together with second harmonic generation experiments, we show that the compound actually adopts a non-centrosymmetric structure. A large anomalous Hall conductivity of 102 {\Omega}^(-1) cm^(-1) at low temperature stems from intrinsic origin, which is larger than any previously reported values in bulk Cr-Te system. In addition, we observe a hump-like feature in the field-dependent Hall resistivity data, resembling a typical topological Hall signal. We demonstrate that the feature is highly tunable and is not related to topological Hall effect even though we observe N\'eel-type skyrmions by Lorentz transmission electron microscopy which is consistent with the non-centrosymmetric structure of the compound., Comment: 26 pages, 8 Figures

We propose a sequential Monte Carlo (SMC) method to efficiently and accurately compute cut-Bayesian posterior quantities of interest, variations of standard Bayesian approaches constructed primarily to account for model misspecification. We prove finite sample concentration bounds for estimators derived from the proposed method along with a linear tempering extension and apply these results to a realistic setting where a computer model is misspecified. We then illustrate the SMC method for inference in a modular chemical reactor example that includes submodels for reaction kinetics, turbulence, mass transfer, and diffusion. The samples obtained are commensurate with a direct-sampling approach that consists of running multiple Markov chains, with computational efficiency gains using the SMC method. Overall, the SMC method presented yields a novel, rigorous approach to computing with cut-Bayesian posterior distributions.

Consider a multi-class labelling problem, where the labels can take values in $[k]$, and a predictor predicts a distribution over the labels. In this work, we study the following foundational question: Are there notions of multi-class calibration that give strong guarantees of meaningful predictions and can be achieved in time and sample complexities polynomial in $k$? Prior notions of calibration exhibit a tradeoff between computational efficiency and expressivity: they either suffer from having sample complexity exponential in $k$, or needing to solve computationally intractable problems, or give rather weak guarantees. Our main contribution is a notion of calibration that achieves all these desiderata: we formulate a robust notion of projected smooth calibration for multi-class predictions, and give new recalibration algorithms for efficiently calibrating predictors under this definition with complexity polynomial in $k$. Projected smooth calibration gives strong guarantees for all downstream decision makers who want to use the predictor for binary classification problems of the form: does the label belong to a subset $T \subseteq [k]$: e.g. is this an image of an animal? It ensures that the probabilities predicted by summing the probabilities assigned to labels in $T$ are close to some perfectly calibrated binary predictor for that task. We also show that natural strengthenings of our definition are computationally hard to achieve: they run into information theoretic barriers or computational intractability. Underlying both our upper and lower bounds is a tight connection that we prove between multi-class calibration and the well-studied problem of agnostic learning in the (standard) binary prediction setting., Comment: In COLT 2024

A central paradigm of non-equilibrium physics concerns the dynamics of heterogeneity and disorder, impacting processes ranging from the behavior of glasses to the emergent functionality of active matter. Understanding these complex mesoscopic systems requires probing the microscopic trajectories associated with irreversible processes, the role of fluctuations and entropy growth, and the timescales on which non-equilibrium responses are ultimately maintained. Approaches that illuminate these processes in model systems may enable a more general understanding of other heterogeneous non-equilibrium phenomena, and potentially define ultimate speed and energy cost limits for information processing technologies. Here, we apply ultrafast single shot x-ray photon correlation spectroscopy to resolve the non-equilibrium, heterogeneous, and irreversible mesoscale dynamics during a light-induced phase transition. This approach defines a new way of capturing the nucleation of the induced phase, the formation of transient mesoscale defects at the boundaries of the nuclei, and the eventual annihilation of these defects, even in systems with complex polarization topologies. A non-equilibrium response spanning >10 orders of magnitude in timescales is observed, with multistep behavior similar to the plateaus observed in supercooled liquids and glasses. We show how the observed time-dependent long-time correlations can be understood in terms of the stochastic dynamics of domain walls, encoded in effective waiting-time distributions with power-law tails. This work defines new possibilities for probing the non-equilibrium and correlated dynamics of disordered and heterogeneous media.

Consider the supervised learning setting where the goal is to learn to predict labels $\mathbf y$ given points $\mathbf x$ from a distribution. An \textit{omnipredictor} for a class $\mathcal L$ of loss functions and a class $\mathcal C$ of hypotheses is a predictor whose predictions incur less expected loss than the best hypothesis in $\mathcal C$ for every loss in $\mathcal L$. Since the work of [GKR+21] that introduced the notion, there has been a large body of work in the setting of binary labels where $\mathbf y \in \{0, 1\}$, but much less is known about the regression setting where $\mathbf y \in [0,1]$ can be continuous. Our main conceptual contribution is the notion of \textit{sufficient statistics} for loss minimization over a family of loss functions: these are a set of statistics about a distribution such that knowing them allows one to take actions that minimize the expected loss for any loss in the family. The notion of sufficient statistics relates directly to the approximate rank of the family of loss functions. Our key technical contribution is a bound of $O(1/\varepsilon^{2/3})$ on the $\epsilon$-approximate rank of convex, Lipschitz functions on the interval $[0,1]$, which we show is tight up to a factor of $\mathrm{polylog} (1/\epsilon)$. This yields improved runtimes for learning omnipredictors for the class of all convex, Lipschitz loss functions under weak learnability assumptions about the class $\mathcal C$. We also give efficient omnipredictors when the loss families have low-degree polynomial approximations, or arise from generalized linear models (GLMs). This translation from sufficient statistics to faster omnipredictors is made possible by lifting the technique of loss outcome indistinguishability introduced by [GKH+23] for Boolean labels to the regression setting.

This paper concerns the development of metatheory for extensible languages. It uses as its starting point a view that programming languages tailored to specific application domains are to be constructed by composing components from an open library of independently-developed extensions to a host language. In the elaboration of this perspective, static analyses (such as typing) and dynamic semantics (such as evaluation) are described via relations whose specifications are distributed across the host language and extensions and are given in a rule-based fashion. Metatheoretic properties, which ensure that static analyses accurately gauge runtime behavior, are represented in this context by formulas over such relations. These properties may be fundamental to the language, introduced by the host language, or they may pertain to analyses introduced by individual extensions. We expose the problem of modular metatheory, i.e., the notion that proofs of relevant properties can be constructed by reasoning independently within each component in the library. To solve this problem, we propose the twin ideas of decomposing proofs around language fragments and of reasoning generically about extensions based on broad, a priori constraints imposed on their behavior. We establish the soundness of these styles of reasoning by showing how complete proofs of the properties can be automatically constructed for any language obtained by composing the independent parts. Mathematical precision is given to our discussions by framing them within a logic that encodes inductive rule-based specifications via least fixed-point definitions. We also sketch the structure of a practical system for metatheoretic reasoning for extensible languages based on the ideas developed.

We present a framework for robots to learn novel visual concepts and tasks via in-situ linguistic interactions with human users. Previous approaches have either used large pre-trained visual models to infer novel objects zero-shot, or added novel concepts along with their attributes and representations to a concept hierarchy. We extend the approaches that focus on learning visual concept hierarchies by enabling them to learn novel concepts and solve unseen robotics tasks with them. To enable a visual concept learner to solve robotics tasks one-shot, we developed two distinct techniques. Firstly, we propose a novel approach, Hi-Viscont(HIerarchical VISual CONcept learner for Task), which augments information of a novel concept to its parent nodes within a concept hierarchy. This information propagation allows all concepts in a hierarchy to update as novel concepts are taught in a continual learning setting. Secondly, we represent a visual task as a scene graph with language annotations, allowing us to create novel permutations of a demonstrated task zero-shot in-situ. We present two sets of results. Firstly, we compare Hi-Viscont with the baseline model (FALCON) on visual question answering(VQA) in three domains. While being comparable to the baseline model on leaf level concepts, Hi-Viscont achieves an improvement of over 9% on non-leaf concepts on average. We compare our model's performance against the baseline FALCON model. Our framework achieves 33% improvements in success rate metric, and 19% improvements in the object level accuracy compared to the baseline model. With both of these results we demonstrate the ability of our model to learn tasks and concepts in a continual learning setting on the robot., Comment: In Proceedings of The 38th Annual AAAI Conference on Artificial Intelligence

Spintronics has emerged as a viable alternative to traditional electronics based technologies in the past few decades. While on one hand, the discovery of topological phases of matter with protected spin-polarized states has opened up exciting prospects, recent revelation of intriguing non-relativistic spin splitting in collinear antiferromagnetic materials with unique symmetries facilitate a wide possibility of realizing both these features simultaneously. In this work, we report the co-existence of these two intriguing properties within a single material: GdAlSi. It crystallizes in a body-centered tetragonal structure with a non-centrosymmetric space group $I4_{1}md$ ($109$), which is confirmed using detailed structural analysis through X-ray diffraction (XRD) and optical second harmonic generation (SHG) measurements. The magnetization data indicates AFM ordering with an ordering temperature ($T_N$) $\sim$ 32 K. Ab-initio calculations reveal GdAlSi to be a collinear antiferromagnetic Weyl semimetal with an unconventional, momentum-dependent spin splitting, also referred to as altermagnet. Angle-resolved photoemission spectroscopy measurements on GdAlSi single crystals subsequently confirm the presence of Fermi arcs, a distinctive hallmark of Weyl semimetals. Electric and magnetic multipole analysis provides a deeper understanding of the symmetry-mediated, momentum-dependent spin splitting, which has strictly non-relativistic origin. To the best of our knowledge, such co-existence of unconventional antiferromagnetic order and non-trivial topology is unprecedented and has never been observed before in a single material, rendering GdAlSi a special and promising candidate material. We propose a device harnessing these features, poised to enable practical and efficient topotronic applications., Comment: 24 pages (14 pages main, 10 pages supplement), 15 figures (7 figures main, 8 figures supplement)