Search

Soft grippers have demonstrated a strong ability to successfully pick and manipulate many objects. A key limitation to their wider adoption is their inability to grasp larger payloads due to objects slipping out of grasps. We have overcome this limitation by introducing a torsionally rigid strain limiting layer (TR-SLL). This reduces out-of-plane bending while maintaining the gripper's softness and in-plane flexibility. We characterize the design space of the strain limiting layer and Handed Shearing Auxetic (HSA) actuators for a soft gripper using simulation and experiment. The inclusion of the TR-SLL with HSAs enables HSA grippers to be made with a single digit. We found that the use of our TR-SLL HSA gripper enabled pinch grasping of payloads over 1 kg. We demonstrate a lifting capacity of 5 kg when loading using the TR-SLL. We also demonstrate a peak pinch grasp force of 5.8 N, and a peak planar caging force of 14.5 N. Finally, we test the TR-SLL gripper on a suite of 43 YCB objects. We show success on 37 objects demonstrating significant capabilities., Comment: 12 pages, 11 figures, paper under review. arXiv admin note: text overlap with arXiv:2311.02454

Torque and continuous rotation are fundamental methods of actuation and manipulation in rigid robots. Soft robot arms use soft materials and structures to mimic the passive compliance of biological arms that bend and extend. This use of compliance prevents soft arms from continuously transmitting and exerting torques to interact with their environment. Here, we show how relying on patterning structures instead of inherent material properties allows soft robotic arms to remain compliant while continuously transmitting torque to their environment. We demonstrate a soft robotic arm made from a pair of mechanical metamaterials that act as compliant constant-velocity joints. The joints are up to 52 times stiffer in torsion than bending and can bend up to 45{\deg}. This robot arm can continuously transmit torque while deforming in all other directions. The arm's mechanical design achieves high motion repeatability (0.4 mm and 0.1{\deg}) when tracking trajectories. We then trained a neural network to learn the inverse kinematics, enabling us to program the arm to complete tasks that are challenging for existing soft robots such as installing light bulbs, fastening bolts, and turning valves. The arm's passive compliance makes it safe around humans and provides a source of mechanical intelligence, enabling it to adapt to misalignment when manipulating objects. This work will bridge the gap between hard and soft robotics with applications in human assistance, warehouse automation, and extreme environments.

Several recent works seek to develop foundation models specifically for medical applications, adapting general-purpose large language models (LLMs) and vision-language models (VLMs) via continued pretraining on publicly available biomedical corpora. These works typically claim that such domain-adaptive pretraining (DAPT) improves performance on downstream medical tasks, such as answering medical licensing exam questions. In this paper, we compare ten public "medical" LLMs and two VLMs against their corresponding base models, arriving at a different conclusion: all medical VLMs and nearly all medical LLMs fail to consistently improve over their base models in the zero-/few-shot prompting and supervised fine-tuning regimes for medical question-answering (QA). For instance, across all tasks and model pairs we consider in the 3-shot setting, medical LLMs only outperform their base models in 22.7% of cases, reach a (statistical) tie in 36.8% of cases, and are significantly worse than their base models in the remaining 40.5% of cases. Our conclusions are based on (i) comparing each medical model head-to-head, directly against the corresponding base model; (ii) optimizing the prompts for each model separately in zero-/few-shot prompting; and (iii) accounting for statistical uncertainty in comparisons. While these basic practices are not consistently adopted in the literature, our ablations show that they substantially impact conclusions. Meanwhile, we find that after fine-tuning on specific QA tasks, medical LLMs can show performance improvements, but the benefits do not carry over to tasks based on clinical notes. Our findings suggest that state-of-the-art general-domain models may already exhibit strong medical knowledge and reasoning capabilities, and offer recommendations to strengthen the conclusions of future studies., Comment: Extended version of EMNLP 2024 paper arXiv:2411.04118. Includes additional results on clinical note QA tasks and supervised fine-tuning evaluations

Several recent works seek to develop foundation models specifically for medical applications, adapting general-purpose large language models (LLMs) and vision-language models (VLMs) via continued pretraining on publicly available biomedical corpora. These works typically claim that such domain-adaptive pretraining (DAPT) improves performance on downstream medical tasks, such as answering medical licensing exam questions. In this paper, we compare seven public "medical" LLMs and two VLMs against their corresponding base models, arriving at a different conclusion: all medical VLMs and nearly all medical LLMs fail to consistently improve over their base models in the zero-/few-shot prompting regime for medical question-answering (QA) tasks. For instance, across the tasks and model pairs we consider in the 3-shot setting, medical LLMs only outperform their base models in 12.1% of cases, reach a (statistical) tie in 49.8% of cases, and are significantly worse than their base models in the remaining 38.2% of cases. Our conclusions are based on (i) comparing each medical model head-to-head, directly against the corresponding base model; (ii) optimizing the prompts for each model separately; and (iii) accounting for statistical uncertainty in comparisons. While these basic practices are not consistently adopted in the literature, our ablations show that they substantially impact conclusions. Our findings suggest that state-of-the-art general-domain models may already exhibit strong medical knowledge and reasoning capabilities, and offer recommendations to strengthen the conclusions of future studies., Comment: This version was published at EMNLP 2024 Main Conference as a Long Paper (Oral). See the extended version (arXiv:2411.08870) for additional results on QA tasks based on clinical notes and evaluations in the supervised fine-tuning regime

While many works have studied statistical data fusion, they typically assume that the various datasets are given in advance. However, in practice, estimation requires difficult data collection decisions like determining the available data sources, their costs, and how many samples to collect from each source. Moreover, this process is often sequential because the data collected at a given time can improve collection decisions in the future. In our setup, given access to multiple data sources and budget constraints, the agent must sequentially decide which data source to query to efficiently estimate a target parameter. We formalize this task using Online Moment Selection, a semiparametric framework that applies to any parameter identified by a set of moment conditions. Interestingly, the optimal budget allocation depends on the (unknown) true parameters. We present two online data collection policies, Explore-then-Commit and Explore-then-Greedy, that use the parameter estimates at a given time to optimally allocate the remaining budget in the future steps. We prove that both policies achieve zero regret (assessed by asymptotic MSE) relative to an oracle policy. We empirically validate our methods on both synthetic and real-world causal effect estimation tasks, demonstrating that the online data collection policies outperform their fixed counterparts.

This document is comprised of a collection of updated preliminary parameters for the key parts of the muon collider. The updated preliminary parameters follow on from the October 2023 Tentative Parameters Report. Particular attention has been given to regions of the facility that are believed to hold greater technical uncertainty in their design and that have a strong impact on the cost and power consumption of the facility. The data is collected from a collaborative spreadsheet and transferred to overleaf.

In this work, we investigate the causal reasoning abilities of large language models (LLMs) through the representative problem of inferring causal relationships from narratives. We find that even state-of-the-art language models rely on unreliable shortcuts, both in terms of the narrative presentation and their parametric knowledge. For example, LLMs tend to determine causal relationships based on the topological ordering of events (i.e., earlier events cause later ones), resulting in lower performance whenever events are not narrated in their exact causal order. Similarly, we demonstrate that LLMs struggle with long-term causal reasoning and often fail when the narratives are long and contain many events. Additionally, we show LLMs appear to rely heavily on their parametric knowledge at the expense of reasoning over the provided narrative. This degrades their abilities whenever the narrative opposes parametric knowledge. We extensively validate these failure modes through carefully controlled synthetic experiments, as well as evaluations on real-world narratives. Finally, we observe that explicitly generating a causal graph generally improves performance while naive chain-of-thought is ineffective. Collectively, our results distill precise failure modes of current state-of-the-art models and can pave the way for future techniques to enhance causal reasoning in LLMs.

Graph neural networks (GNNs) are the dominant approach to solving machine learning problems defined over graphs. Despite much theoretical and empirical work in recent years, our understanding of finer-grained aspects of architectural design for GNNs remains impoverished. In this paper, we consider the benefits of architectures that maintain and update edge embeddings. On the theoretical front, under a suitable computational abstraction for a layer in the model, as well as memory constraints on the embeddings, we show that there are natural tasks on graphical models for which architectures leveraging edge embeddings can be much shallower. Our techniques are inspired by results on time-space tradeoffs in theoretical computer science. Empirically, we show architectures that maintain edge embeddings almost always improve on their node-based counterparts -- frequently significantly so in topologies that have ``hub'' nodes., Comment: 25 pages, 2 figures

Fairness metrics are a core tool in the fair machine learning literature (FairML), used to determine that ML models are, in some sense, "fair". Real-world data, however, are typically plagued by various measurement biases and other violated assumptions, which can render fairness assessments meaningless. We adapt tools from causal sensitivity analysis to the FairML context, providing a general framework which (1) accommodates effectively any combination of fairness metric and bias that can be posed in the "oblivious setting"; (2) allows researchers to investigate combinations of biases, resulting in non-linear sensitivity; and (3) enables flexible encoding of domain-specific constraints and assumptions. Employing this framework, we analyze the sensitivity of the most common parity metrics under 3 varieties of classifier across 14 canonical fairness datasets. Our analysis reveals the striking fragility of fairness assessments to even minor dataset biases. We show that causal sensitivity analysis provides a powerful and necessary toolkit for gauging the informativeness of parity metric evaluations. Our repository is available here: https://github.com/Jakefawkes/fragile_fair., Comment: Published at Neurips 2024 in the Dataset and Benchmarks Track

As recommender systems become widely deployed in different domains, they increasingly influence their users' beliefs and preferences. Auditing recommender systems is crucial as it not only ensures the continuous improvement of recommendation algorithms but also safeguards against potential issues like biases and ethical concerns. In this paper, we view recommender system auditing from a causal lens and provide a general recipe for defining auditing metrics. Under this general causal auditing framework, we categorize existing auditing metrics and identify gaps in them -- notably, the lack of metrics for auditing user agency while accounting for the multi-step dynamics of the recommendation process. We leverage our framework and propose two classes of such metrics:future- and past-reacheability and stability, that measure the ability of a user to influence their own and other users' recommendations, respectively. We provide both a gradient-based and a black-box approach for computing these metrics, allowing the auditor to compute them under different levels of access to the recommender system. In our experiments, we demonstrate the efficacy of methods for computing the proposed metrics and inspect the design of recommender systems through these proposed metrics., Comment: 28 pages

The International Muon Collider Collaboration (IMCC) [1] was established in 2020 following the recommendations of the European Strategy for Particle Physics (ESPP) and the implementation of the European Strategy for Particle Physics-Accelerator R&D Roadmap by the Laboratory Directors Group [2], hereinafter referred to as the the European LDG roadmap. The Muon Collider Study (MuC) covers the accelerator complex, detectors and physics for a future muon collider. In 2023, European Commission support was obtained for a design study of a muon collider (MuCol) [3]. This project started on 1st March 2023, with work-packages aligned with the overall muon collider studies. In preparation of and during the 2021-22 U.S. Snowmass process, the muon collider project parameters, technical studies and physics performance studies were performed and presented in great detail. Recently, the P5 panel [4] in the U.S. recommended a muon collider R&D, proposed to join the IMCC and envisages that the U.S. should prepare to host a muon collider, calling this their "muon shot". In the past, the U.S. Muon Accelerator Programme (MAP) [5] has been instrumental in studies of concepts and technologies for a muon collider., Comment: This document summarises the International Muon Collider Collaboration (IMCC) progress and status of the Muon Collider R&D programme

Foams are versatile by nature and ubiquitous in a wide range of applications, including padding, insulation, and acoustic dampening. Previous work established that foams 3D printed via Viscous Thread Printing (VTP) can in principle combine the flexibility of 3D printing with the mechanical properties of conventional foams. However, the generality of prior work is limited due to the lack of predictable process-property relationships. In this work, we utilize a self-driving lab that combines automated experimentation with machine learning to identify a processing subspace in which dimensionally consistent materials are produced using VTP with spatially programmable mechanical properties. In carrying out this process, we discover an underlying self-stabilizing characteristic of VTP layer thickness, an important feature for its extension to new materials and systems. Several complex exemplars are constructed to illustrate the newly enabled capabilities of foams produced via VTP, including 1D gradient rectangular slabs, 2D localized stiffness zones on an insole orthotic and living hinges, and programmed 3D deformation via a cable driven humanoid hand. Predictive mapping models are developed and validated for both thermoplastic polyurethane (TPU) and polylactic acid (PLA) filaments, suggesting the ability to train a model for any material suitable for material extrusion (ME) 3D printing., Comment: 23 pages, 7 figures

We develop static and dynamic approaches for hedging of the impermanent loss (IL) of liquidity provision (LP) staked at Decentralised Exchanges (DEXes) which employ Uniswap V2 and V3 protocols. We provide detailed definitions and formulas for computing the IL to unify different definitions occurring in the existing literature. We show that the IL can be seen a contingent claim with a non-linear payoff for a fixed maturity date. Thus, we introduce the contingent claim termed as IL protection claim which delivers the negative of IL payoff at the maturity date. We apply arbitrage-based methods for valuation and risk management of this claim. First, we develop the static model-independent replication method for the valuation of IL protection claim using traded European vanilla call and put options. We extend and generalize an existing method to show that the IL protection claim can be hedged perfectly with options if there is a liquid options market. Second, we develop the dynamic model-based approach for the valuation and hedging of IL protection claims under a risk-neutral measure. We derive analytic valuation formulas using a wide class of price dynamics for which the characteristic function is available under the risk-neutral measure. As base cases, we derive analytic valuation formulas for IL protection claim under the Black-Scholes-Merton model and the log-normal stochastic volatility model. We finally discuss estimation of risk-reward of LP staking using our results., Comment: 34 pages, 8 figures

Mechanical extremal materials, a class of metamaterials that exist at the bounds of elastic theory, possess the extraordinary capability to engineer any desired elastic behavior by harnessing mechanical zero modes -- deformation modes that demand minimal or no elastic energy. However, the potential for arbitrary construction and reprogramming of metamaterials remains largely unrealized, primarily due to significant challenges in qualitatively transforming zero modes within the confines of existing metamaterial design frameworks. In this work, we show a method for explicitly defining and in situ reprogramming zero modes of two-dimensional extremal materials by employing straight-line mechanisms (SLMs) and planar symmetry, which prescribe and coordinate the zero modes, respectively. We design, test, and reprogram centimeter-scale isotropic, orthotropic, and chiral extremal materials by reorienting the SLMs in place, enabling these materials to smoothly and reversibly interpolate between extremal modalities (e.g., unimode to bimode) and material properties (e.g., negative to positive Poisson's ratios) without changing the metamaterial's global structure. Our methodology provides a straightforward and explicit strategy for the design and tuning of all varieties of two-dimensional extremal materials, enabling arbitrary and dynamic mechanical metamaterial construction to completely cover the gamut of elastic properties., Comment: 14 pages, 4 figures

In this paper, we demonstrate a surprising capability of large language models (LLMs): given only input feature names and a description of a prediction task, they are capable of selecting the most predictive features, with performance rivaling the standard tools of data science. Remarkably, these models exhibit this capacity across various query mechanisms. For example, we zero-shot prompt an LLM to output a numerical importance score for a feature (e.g., "blood pressure") in predicting an outcome of interest (e.g., "heart failure"), with no additional context. In particular, we find that the latest models, such as GPT-4, can consistently identify the most predictive features regardless of the query mechanism and across various prompting strategies. We illustrate these findings through extensive experiments on real-world data, where we show that LLM-based feature selection consistently achieves strong performance competitive with data-driven methods such as the LASSO, despite never having looked at the downstream training data. Our findings suggest that LLMs may be useful not only for selecting the best features for training but also for deciding which features to collect in the first place. This could potentially benefit practitioners in domains like healthcare, where collecting high-quality data comes at a high cost., Comment: Preprint

Colloquially speaking, image generation models based upon diffusion processes are frequently said to exhibit "hallucinations," samples that could never occur in the training data. But where do such hallucinations come from? In this paper, we study a particular failure mode in diffusion models, which we term mode interpolation. Specifically, we find that diffusion models smoothly "interpolate" between nearby data modes in the training set, to generate samples that are completely outside the support of the original training distribution; this phenomenon leads diffusion models to generate artifacts that never existed in real data (i.e., hallucinations). We systematically study the reasons for, and the manifestation of this phenomenon. Through experiments on 1D and 2D Gaussians, we show how a discontinuous loss landscape in the diffusion model's decoder leads to a region where any smooth approximation will cause such hallucinations. Through experiments on artificial datasets with various shapes, we show how hallucination leads to the generation of combinations of shapes that never existed. Finally, we show that diffusion models in fact know when they go out of support and hallucinate. This is captured by the high variance in the trajectory of the generated sample towards the final few backward sampling process. Using a simple metric to capture this variance, we can remove over 95% of hallucinations at generation time while retaining 96% of in-support samples. We conclude our exploration by showing the implications of such hallucination (and its removal) on the collapse (and stabilization) of recursive training on synthetic data with experiments on MNIST and 2D Gaussians dataset. We release our code at https://github.com/locuslab/diffusion-model-hallucination., Comment: Additional results on real datasets

Recent advancements in general-purpose AI have highlighted the importance of guiding AI systems towards the intended goals, ethical principles, and values of individuals and groups, a concept broadly recognized as alignment. However, the lack of clarified definitions and scopes of human-AI alignment poses a significant obstacle, hampering collaborative efforts across research domains to achieve this alignment. In particular, ML- and philosophy-oriented alignment research often views AI alignment as a static, unidirectional process (i.e., aiming to ensure that AI systems' objectives match humans) rather than an ongoing, mutual alignment problem. This perspective largely neglects the long-term interaction and dynamic changes of alignment. To understand these gaps, we introduce a systematic review of over 400 papers published between 2019 and January 2024, spanning multiple domains such as Human-Computer Interaction (HCI), Natural Language Processing (NLP), Machine Learning (ML). We characterize, define and scope human-AI alignment. From this, we present a conceptual framework of "Bidirectional Human-AI Alignment" to organize the literature from a human-centered perspective. This framework encompasses both 1) conventional studies of aligning AI to humans that ensures AI produces the intended outcomes determined by humans, and 2) a proposed concept of aligning humans to AI, which aims to help individuals and society adjust to AI advancements both cognitively and behaviorally. Additionally, we articulate the key findings derived from literature analysis, including literature gaps and trends, human values, and interaction techniques. To pave the way for future studies, we envision three key challenges and give recommendations for future research., Comment: proposing "bidirectional human-AI alignment" framework after a systematic review of over 400 alignment papers

Recent advancements in large language models (LLMs) have considerably advanced the capabilities of summarization systems. However, they continue to face concerns about hallucinations. While prior work has evaluated LLMs extensively in news domains, most evaluation of dialogue summarization has focused on BART-based models, leaving a gap in our understanding of their faithfulness. Our work benchmarks the faithfulness of LLMs for dialogue summarization, using human annotations and focusing on identifying and categorizing span-level inconsistencies. Specifically, we focus on two prominent LLMs: GPT-4 and Alpaca-13B. Our evaluation reveals subtleties as to what constitutes a hallucination: LLMs often generate plausible inferences, supported by circumstantial evidence in the conversation, that lack direct evidence, a pattern that is less prevalent in older models. We propose a refined taxonomy of errors, coining the category of "Circumstantial Inference" to bucket these LLM behaviors and release the dataset. Using our taxonomy, we compare the behavioral differences between LLMs and older fine-tuned models. Additionally, we systematically assess the efficacy of automatic error detection methods on LLM summaries and find that they struggle to detect these nuanced errors. To address this, we introduce two prompt-based approaches for fine-grained error detection that outperform existing metrics, particularly for identifying "Circumstantial Inference.", Comment: Accepted at ACL 2024

In this article, a geometric approach to incorporating investor views in portfolio construction is presented. In particular, the proposed approach utilizes the notion of generalized Wasserstein barycenter (GWB) to combine the statistical information about asset returns with investor views to obtain an updated estimate of the asset drifts and covariance, which are then fed into a mean-variance optimizer as inputs. Quantitative comparisons of the proposed geometric approach with the conventional Black-Litterman model (and a closely related variant) are presented. The proposed geometric approach provides investors with more flexibility in specifying their confidence in their views than conventional Black-Litterman model-based approaches. The geometric approach also rewards the investors more for making correct decisions than conventional BL based approaches. We provide empirical and theoretical justifications for our claim., Comment: Abstract Page + TOC + 35 pages, 4 Figures, 4 Tables

This paper introduces a model theory for resolution on Higher Order Hereditarily Harrop formulae (HOHH), the logic underlying the Lambda-Prolog programming language, and proves soundness and completeness of resolution. The semantics and the proof of completeness of the formal system is shown in several ways, suitably adapted to deal with the impredicativity of higher-order logic, which rules out definitions of truth based on induction on formula structure. First, we use the least fixed point of a certain operator on interpretations, in the style of Apt and Van Emden, Then a constructive completeness theorem is given using a proof theoretic variant of the Lindenbaum algebra, which also contains a new approach to establishing cut-elimination., Comment: 64 pages, 2 Figures

Large language models (LLMs) trained on web-scale datasets raise substantial concerns regarding permissible data usage. One major question is whether these models "memorize" all their training data or they integrate many data sources in some way more akin to how a human would learn and synthesize information. The answer hinges, to a large degree, on how we define memorization. In this work, we propose the Adversarial Compression Ratio (ACR) as a metric for assessing memorization in LLMs. A given string from the training data is considered memorized if it can be elicited by a prompt (much) shorter than the string itself -- in other words, if these strings can be "compressed" with the model by computing adversarial prompts of fewer tokens. The ACR overcomes the limitations of existing notions of memorization by (i) offering an adversarial view of measuring memorization, especially for monitoring unlearning and compliance; and (ii) allowing for the flexibility to measure memorization for arbitrary strings at a reasonably low compute. Our definition serves as a practical tool for determining when model owners may be violating terms around data usage, providing a potential legal tool and a critical lens through which to address such scenarios., Comment: https://locuslab.github.io/acr-memorization

Trained models are often composed with post-hoc transforms such as temperature scaling (TS), ensembling and stochastic weight averaging (SWA) to improve performance, robustness, uncertainty estimation, etc. However, such transforms are typically applied only after the base models have already been finalized by standard means. In this paper, we challenge this practice with an extensive empirical study. In particular, we demonstrate a phenomenon that we call post-hoc reversal, where performance trends are reversed after applying post-hoc transforms. This phenomenon is especially prominent in high-noise settings. For example, while base models overfit badly early in training, both ensembling and SWA favor base models trained for more epochs. Post-hoc reversal can also prevent the appearance of double descent and mitigate mismatches between test loss and test error seen in base models. Preliminary analyses suggest that these transforms induce reversal by suppressing the influence of mislabeled examples, exploiting differences in their learning dynamics from those of clean examples. Based on our findings, we propose post-hoc selection, a simple technique whereby post-hoc metrics inform model development decisions such as early stopping, checkpointing, and broader hyperparameter choices. Our experiments span real-world vision, language, tabular and graph datasets. On an LLM instruction tuning dataset, post-hoc selection results in >1.5x MMLU improvement compared to naive selection., Comment: accepted at NeurIPS 2024; v2 adds an intuitions section

Vision-language models (VLMs) are trained for thousands of GPU hours on carefully curated web datasets. In recent times, data curation has gained prominence with several works developing strategies to retain 'high-quality' subsets of 'raw' scraped data. For instance, the LAION public dataset retained only 10% of the total crawled data. However, these strategies are typically developed agnostic of the available compute for training. In this paper, we first demonstrate that making filtering decisions independent of training compute is often suboptimal: the limited high-quality data rapidly loses its utility when repeated, eventually requiring the inclusion of 'unseen' but 'lower-quality' data. To address this quality-quantity tradeoff ($\texttt{QQT}$), we introduce neural scaling laws that account for the non-homogeneous nature of web data, an angle ignored in existing literature. Our scaling laws (i) characterize the $\textit{differing}$ 'utility' of various quality subsets of web data; (ii) account for how utility diminishes for a data point at its 'nth' repetition; and (iii) formulate the mutual interaction of various data pools when combined, enabling the estimation of model performance on a combination of multiple data pools without ever jointly training on them. Our key message is that data curation $\textit{cannot}$ be agnostic of the total compute that a model will be trained for. Our scaling laws allow us to curate the best possible pool for achieving top performance on Datacomp at various compute budgets, carving out a pareto-frontier for data curation. Code is available at https://github.com/locuslab/scaling_laws_data_filtering., Comment: Published at CVPR 2024

Many definitions of fairness or inequity involve unobservable causal quantities that cannot be directly estimated without strong assumptions. For instance, it is particularly difficult to estimate notions of fairness that rely on hard-to-measure concepts such as risk (e.g., quantifying whether patients at the same risk level have equal probability of treatment, regardless of group membership). Such measurements of risk can be accurately obtained when no unobserved confounders have jointly influenced past decisions and outcomes. However, in the real world, this assumption rarely holds. In this paper, we show that, surprisingly, one can still compute meaningful bounds on treatment rates for high-risk individuals (i.e., conditional on their true, \textit{unobserved} negative outcome), even when entirely eliminating or relaxing the assumption that we observe all relevant risk factors used by decision makers. We use the fact that in many real-world settings (e.g., the release of a new treatment) we have data from prior to any allocation to derive unbiased estimates of risk. This result enables us to audit unfair outcomes of existing decision-making systems in a principled manner. We demonstrate the effectiveness of our framework with a real-world study of Paxlovid allocation, provably identifying that observed racial inequity cannot be explained by unobserved confounders of the same strength as important observed covariates., Comment: AISTATS 2024

An intriguing link between a wide range of problems occurring in physics and financial engineering is presented. These problems include the evolution of small perturbations of linear flows in hydrodynamics, the movements of particles in random fields described by the Kolmogorov and Klein-Kramers equations, the Ornstein-Uhlenbeck and Feller processes, and their generalizations. They are reduced to affine differential and pseudo-differential equations and solved in a unified way by using Kelvin waves and developing a comprehensive math framework for calculating transition probabilities and expectations. Kelvin waves are instrumental for studying the well-known Black-Scholes, Heston, and Stein-Stein models and more complex path-dependent volatility models, as well as the pricing of Asian options, volatility and variance swaps, bonds, and bond options. Kelvin waves help to solve several cutting-edge problems, including hedging the impermanent loss of Automated Market Makers for cryptocurrency trading. This title is also available as Open Access on Cambridge Core., Comment: 123 pages, 14 figures. arXiv admin note: text overlap with arXiv:2309.04547

This work considers the nodal finite element approximation of peridynamics, in which the nodal displacements satisfy the peridynamics equation at each mesh node. For the nonlinear bond-based peridynamics model, it is shown that, under the suitable assumptions on an exact solution, the discretized solution associated with the central-in-time and nodal finite element discretization converges to the exact solution in $L^2$ norm at the rate $C_1 \Delta t + C_2 h^2/\epsilon^2$. Here, $\Delta t$, $h$, and $\epsilon$ are time step size, mesh size, and the size of the horizon or nonlocal length scale, respectively. Constants $C_1$ and $C_2$ are independent of $h$ and $\Delta t$ and depend on norms of the solution and nonlocal length scale. Several numerical examples involving pre-crack, void, and notch are considered, and the efficacy of the proposed nodal finite element discretization is analyzed., Comment: 35 pages, 22 figures

LLMs can generate factually incorrect statements even when provided access to reference documents. Such errors can be dangerous in high-stakes applications (e.g., document-grounded QA for healthcare or finance). We present GenAudit -- a tool intended to assist fact-checking LLM responses for document-grounded tasks. GenAudit suggests edits to the LLM response by revising or removing claims that are not supported by the reference document, and also presents evidence from the reference for facts that do appear to have support. We train models to execute these tasks, and design an interactive interface to present suggested edits and evidence to users. Comprehensive evaluation by human raters shows that GenAudit can detect errors in 8 different LLM outputs when summarizing documents from diverse domains. To ensure that most errors are flagged by the system, we propose a method that can increase the error recall while minimizing impact on precision. We release our tool (GenAudit) and fact-checking model for public use., Comment: Code and models available at https://genaudit.org

The acquisition of large-scale, precisely labeled datasets for person re-identification (ReID) poses a significant challenge. Weakly supervised ReID has begun to address this issue, although its performance lags behind fully supervised methods. In response, we introduce Contrastive Multiple Instance Learning (CMIL), a novel framework tailored for more effective weakly supervised ReID. CMIL distinguishes itself by requiring only a single model and no pseudo labels while leveraging contrastive losses -- a technique that has significantly enhanced traditional ReID performance yet is absent in all prior MIL-based approaches. Through extensive experiments and analysis across three datasets, CMIL not only matches state-of-the-art performance on the large-scale SYSU-30k dataset with fewer assumptions but also consistently outperforms all baselines on the WL-market1501 and Weakly Labeled MUddy racer re-iDentification dataset (WL-MUDD) datasets. We introduce and release the WL-MUDD dataset, an extension of the MUDD dataset featuring naturally occurring weak labels from the real-world application at PerformancePhoto.co. All our code and data are accessible at https://drive.google.com/file/d/1rjMbWB6m-apHF3Wg_cfqc8QqKgQ21AsT/view?usp=drive_link.

Despite significant progress in optical character recognition (OCR) and computer vision systems, robustly recognizing text and identifying people in images taken in unconstrained \emph{in-the-wild} environments remain an ongoing challenge. However, such obstacles must be overcome in practical applications of vision systems, such as identifying racers in photos taken during off-road racing events. To this end, we introduce two new challenging real-world datasets - the off-road motorcycle Racer Number Dataset (RND) and the Muddy Racer re-iDentification Dataset (MUDD) - to highlight the shortcomings of current methods and drive advances in OCR and person re-identification (ReID) under extreme conditions. These two datasets feature over 6,300 images taken during off-road competitions which exhibit a variety of factors that undermine even modern vision systems, namely mud, complex poses, and motion blur. We establish benchmark performance on both datasets using state-of-the-art models. Off-the-shelf models transfer poorly, reaching only 15% end-to-end (E2E) F1 score on text spotting, and 33% rank-1 accuracy on ReID. Fine-tuning yields major improvements, bringing model performance to 53% F1 score for E2E text spotting and 79% rank-1 accuracy on ReID, but still falls short of good performance. Our analysis exposes open problems in real-world OCR and ReID that necessitate domain-targeted techniques. With these datasets and analysis of model limitations, we aim to foster innovations in handling real-world conditions like mud and complex poses to drive progress in robust computer vision. All data was sourced from PerformancePhoto.co, a website used by professional motorsports photographers, racers, and fans. The top-performing text spotting and ReID models are deployed on this platform to power real-time race photo search., Comment: arXiv admin note: substantial text overlap with arXiv:2311.09256

Recent work has shown that large language models (LLMs) are capable of generating summaries zero-shot (i.e., without explicit supervision) that, under human assessment, are often comparable or even preferred to manually composed reference summaries. However, this prior work has focussed almost exclusively on evaluating news article summarization. How do zero-shot summarizers perform in other (potentially more specialized) domains? In this work we evaluate zero-shot generated summaries across specialized domains including biomedical articles, and legal bills (in addition to standard news benchmarks for reference). We focus especially on the factuality of outputs. We acquire annotations from domain experts to identify inconsistencies in summaries and systematically categorize these errors. We analyze whether the prevalence of a given domain in the pretraining corpus affects extractiveness and faithfulness of generated summaries of articles in this domain. We release all collected annotations to facilitate additional research toward measuring and realizing factually accurate summarization, beyond news articles. The dataset can be downloaded from https://github.com/sanjanaramprasad/zero_shot_faceval_domains

Personalized large language models (LLMs) are designed to tailor responses to individual user preferences. While Reinforcement Learning from Human Feedback (RLHF) is a commonly used framework for aligning LLMs with human preferences, vanilla RLHF assumes that all human preferences share the same distribution, preventing fine-tuned LLMs from generating personalized content when user preferences are diverse. In this work, we propose Personalized-RLHF (P-RLHF), an efficient framework that utilizes a lightweight user model to capture individual user preferences and jointly learns the user model and the personalized LLM from human feedback. P-RLHF exhibits the following three characteristics: (1) It enables an LLM to generate personalized content and scale efficiently with growing number of users. (2) It handles both explicit user preferences described as textual input and implicit user preferences encoded in the feedback data. (3) It eliminates the need for users to fully articulate their preferences, which are normally needed for prompting LLMs to generate personalized content yet are often impractical to obtain in real-world scenarios. Our experimental results show that personalized LLMs trained using P-RLHF generate responses that are more closely aligned with individual user preferences, outperforming vanilla, non-personalized RLHF and prompting-based personalization approaches across different tasks. We opensource our code at https://github.com/HumainLab/Personalized_RLHF.

Partition of unity methods (PUM) are of domain decomposition type and provide the opportunity for multiscale and multiphysics numerical modeling. Within the PUM global-local enrichment scheme [1, 2] different physical models can exist to capture multiscale behavior. For instance, we consider classical linear elasticity globally and local zones where fractures occur. The elastic fields of the undamaged media provide appropriate boundary data for local PD simulations on a subdomain containing the crack tip to grow the crack path. Once the updated crack path is found, the elastic field in the body and surrounding the crack is updated using PUM basis with appropriate enrichment near the crack. The subdomain for the PD simulation is chosen to include the current crack tip as well as nearby features that will influence crack growth. This paper is part II of this series and validates the combined PD/PUM simulator against the experimental results presented in [3]. The presented results show that we can attain good agreement between experimental and simulation data with a local PD subdomain that is moving with the crack tip and adaptively chosen size.

In response to rising concerns surrounding the safety, security, and trustworthiness of Generative AI (GenAI) models, practitioners and regulators alike have pointed to AI red-teaming as a key component of their strategies for identifying and mitigating these risks. However, despite AI red-teaming's central role in policy discussions and corporate messaging, significant questions remain about what precisely it means, what role it can play in regulation, and how it relates to conventional red-teaming practices as originally conceived in the field of cybersecurity. In this work, we identify recent cases of red-teaming activities in the AI industry and conduct an extensive survey of relevant research literature to characterize the scope, structure, and criteria for AI red-teaming practices. Our analysis reveals that prior methods and practices of AI red-teaming diverge along several axes, including the purpose of the activity (which is often vague), the artifact under evaluation, the setting in which the activity is conducted (e.g., actors, resources, and methods), and the resulting decisions it informs (e.g., reporting, disclosure, and mitigation). In light of our findings, we argue that while red-teaming may be a valuable big-tent idea for characterizing GenAI harm mitigations, and that industry may effectively apply red-teaming and other strategies behind closed doors to safeguard AI, gestures towards red-teaming (based on public definitions) as a panacea for every possible risk verge on security theater. To move toward a more robust toolbox of evaluations for generative AI, we synthesize our recommendations into a question bank meant to guide and scaffold future AI red-teaming practices., Comment: AIES 2024

Predictive risk models in the public sector are commonly developed using administrative data that is more complete for subpopulations that more greatly rely on public services. In the United States, for instance, information on health care utilization is routinely available to government agencies for individuals supported by Medicaid and Medicare, but not for the privately insured. Critiques of public sector algorithms have identified such differential feature under-reporting as a driver of disparities in algorithmic decision-making. Yet this form of data bias remains understudied from a technical viewpoint. While prior work has examined the fairness impacts of additive feature noise and features that are clearly marked as missing, the setting of data missingness absent indicators (i.e. differential feature under-reporting) has been lacking in research attention. In this work, we present an analytically tractable model of differential feature under-reporting which we then use to characterize the impact of this kind of data bias on algorithmic fairness. We demonstrate how standard missing data methods typically fail to mitigate bias in this setting, and propose a new set of methods specifically tailored to differential feature under-reporting. Our results show that, in real world data settings, under-reporting typically leads to increasing disparities. The proposed solution methods show success in mitigating increases in unfairness., Comment: ACM Conference on Fairness, Accountability, and Transparency (FAccT 2024)

Large language models trained on massive corpora of data from the web can memorize and reproduce sensitive or private data raising both legal and ethical concerns. Unlearning, or tuning models to forget information present in their training data, provides us with a way to protect private data after training. Although several methods exist for such unlearning, it is unclear to what extent they result in models equivalent to those where the data to be forgotten was never learned in the first place. To address this challenge, we present TOFU, a Task of Fictitious Unlearning, as a benchmark aimed at helping deepen our understanding of unlearning. We offer a dataset of 200 diverse synthetic author profiles, each consisting of 20 question-answer pairs, and a subset of these profiles called the forget set that serves as the target for unlearning. We compile a suite of metrics that work together to provide a holistic picture of unlearning efficacy. Finally, we provide a set of baseline results from existing unlearning algorithms. Importantly, none of the baselines we consider show effective unlearning motivating continued efforts to develop approaches for unlearning that effectively tune models so that they truly behave as if they were never trained on the forget data at all., Comment: https://locuslab.github.io/tofu/

A nonlocal model of peridynamic type for dynamic brittle damage is introduced consisting of two phases, one elastic and the other inelastic. Evolution from the elastic to the inelastic phase depends on material strength. Existence and uniqueness of the displacement-failure set pair follow from the initial value problem. The displacement-failure pair satisfies energy balance. The length of nonlocality $\epsilon$ is taken to be small relative to the domain in $\mathbb{R}^d$, $d=2,3$. The new nonlocal model delivers a two point strain evolution on a subset of $\mathbb{R}^d\times\mathbb{R}^d$. This evolution provides an energy that interpolates between volume energy corresponding to elastic behavior and surface energy corresponding to failure. In general the deformation energy resulting in material failure over a region $R$ is given by a $d-1$ dimensional integral that is uniformly bounded as $\epsilon\rightarrow 0$. For fixed $\epsilon$, the failure energy is nonzero for $d-1$ dimensional regions $R$ associated with flat crack surfaces. This failure energy is the Griffith fracture energy given by the energy release rate multiplied by area for $d=3$ (or length for $d=2$). The nonlocal field theory is shown to recover a solution of Naiver's equation outside a propagating flat traction free crack in the limit of vanishing spatial nonlocality. Simulations illustrate fracture evolution through generation of an internal traction free boundary as a wake left behind a moving strain concentration. Crack paths are seen to follow a maximal strain energy density criterion., Comment: Updated the title, simplified the discussion on calibration