The opioid crisis represents a significant moment in public health that reveals systemic shortcomings across regulatory systems, healthcare practices, corporate governance, and public policy. Analyzing how these interconnected systems simultaneously failed to protect public health requires innovative analytic approaches for exploring the vast amounts of data and documents disclosed in the UCSF-JHU Opioid Industry Documents Archive (OIDA). The complexity, multimodal nature, and specialized characteristics of these healthcare-related legal and corporate documents necessitate more advanced methods and models tailored to specific data types and detailed annotations, ensuring the precision and professionalism in the analysis. In this paper, we tackle this challenge by organizing the original dataset according to document attributes and constructing a benchmark with 400k training documents and 10k for testing. From each document, we extract rich multimodal information—including textual content, visual elements, and layout structures—to capture a comprehensive range of features. Using multiple AI models, we then generate a large-scale dataset comprising 360k training QA pairs and 10k testing QA pairs. Building on this foundation, we develop domain-specific multimodal Large Language Models (LLMs) and explore the impact of multimodal inputs on task performance. To further enhance response accuracy, we incorporate historical QA pairs as contextual grounding for answering current queries. Additionally, we incorporate page references within the answers and introduce an importance-based page classifier, further improving the precision and relevance of the information provided. Preliminary results indicate the improvements with our AI assistant in document information extraction and question-answering tasks.

Large language models (LLMs) enabled conversational robots to shift toward free-form interaction. However, without context-specific adaptation, generic LLM outputs can be ineffective or inappropriate. This adaptation is often attempted through prompt engineering, which is non-intuitive and tedious. Moreover, predominant design practice in HRI relies on impression-based, trial-and-error refinement without structured methods or tools, making the process inefficient and inconsistent. To address this, we present AI-Aided Conversation Engine (ACE) to support deliberate design of human-robot conversations with three key innovations: 1) an LLM-powered voice agent that scaffolds initial prompt creation to overcome the "blank page problem," 2) an annotation interface that enables the collection of granular and grounded feedback on conversational transcripts, and 3) using LLMs to translate user feedback into prompt refinements. We evaluated ACE through two user studies, examining both designs' experience and end users' interactions with robots designed using ACE. Results show that ACE facilitates the creation of robot behavior prompts with greater clarity and specificity, and that the prompts generated with ACE lead to higher-quality human-robot conversational interactions.

By leveraging the representation power of deep neural networks, neural upper confidence bound (UCB) algorithms have shown success in contextual bandits. To further balance the exploration and exploitation, we propose Neural-$\sigma^2$-LinearUCB, a variance-aware algorithm that utilizes $\sigma^2_t$, i.e., an upper bound of the reward noise variance at round $t$, to enhance the uncertainty quantification quality of the UCB, resulting in a regret performance improvement. We provide an oracle version for our algorithm characterized by an oracle variance upper bound $\sigma^2_t$ and a practical version with a novel estimation for this variance bound. Theoretically, we provide rigorous regret analysis for both versions and prove that our oracle algorithm achieves a better regret guarantee than other neural-UCB algorithms in the neural contextual bandits setting. Empirically, our practical method enjoys a similar computational efficiency, while outperforming state-of-the-art techniques by having a better calibration and lower regret across multiple standard settings, including on the synthetic, UCI, MNIST, and CIFAR-10 datasets.

Responsibly deploying artificial intelligence (AI) / machine learning (ML) systems in high-stakes settings arguably requires not only proof of system reliability, but also continual, post-deployment monitoring to quickly detect and address any unsafe behavior. Methods for nonparametric sequential testing—especially conformal test martingales (CTMs) and anytime-valid inference—offer promising tools for this monitoring task. However, existing approaches are restricted to monitoring limited hypothesis classes or “alarm criteria” (e.g., detecting data shifts that violate certain exchangeability or IID assumptions), do not allow for online adaptation in response to shifts, and/or cannot diagnose the cause of degradation or alarm. In this paper, we address these limitations by proposing a weighted generalization of conformal test martingales (WCTMs), which lay a theoretical foundation for online monitoring for any unexpected changepoints in the data distribution while controlling false-alarms. For practical applications, we propose specific WCTM algorithms that adapt online to mild covariate shifts (in the marginal input distribution), quickly detect harmful shifts, and diagnose those harmful shifts as concept shifts (in the conditional label distribution) or extreme (out-of-support) covariate shifts that cannot be easily adapted to. On real-world datasets, we demonstrate improved performance relative to state-of-the-art baselines.

Language model outputs are not always reliable, thus prompting research into how to adapt model responses based on uncertainty. Common approaches include: abstention, where models refrain from generating responses when uncertain; and linguistic calibration, where models hedge their statements using uncertainty quantifiers. However, abstention can withhold valuable information, while linguistically calibrated responses are often challenging to leverage in downstream tasks. We propose a unified view, Conformal Linguistic Calibration (CLC), which reinterprets linguistic calibration as answer set prediction. First we present a framework connecting abstention and linguistic calibration through the lens of linguistic pragmatics. We then describe an implementation of CLC that allows for controlling the level of imprecision in model responses. Results demonstrate our method produces calibrated outputs with conformal guarantees on factual accuracy. Further, our approach enables fine-tuning models to perform uncertainty-aware adaptive claim rewriting, offering a controllable balance between factuality and specificity.

Zhengping Jiang, Yining Lu, Hanjie Chen, Daniel Khashabi, Benjamin Van Durme, Anqi Liu. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Research in document image understanding is hindered by limited high-quality document data. To address this, we introduce ADOPD, a comprehensive dataset for document page decomposition. ADOPD stands out with its data-driven approach for document taxonomy discovery during data collection, complemented by dense annotations. Our approach integrates large-scale pretrained models with a human-in-the-loop process to guarantee diversity and balance in the resulting data collection. Leveraging our data-driven document taxonomy, we collect and densely annotate document images, addressing four document image understanding tasks: Doc2Mask, Doc2Box, Doc2Tag, and Doc2Seq. Specifically, for each image, the annotations include human-labeled entity masks, text bounding boxes, as well as automatically generated tags and captions that have been manually cleaned. We conduct comprehensive experimental analyses to validate our data and assess the four tasks using various models. We envision ADOPD as a foundational dataset with the potential to drive future research in document understanding.

As artificial intelligence (AI) / machine learning (ML) gain widespread adoption, practitioners are increasingly seeking means to quantify and control the risk these systems incur. This challenge is especially salient when such systems have autonomy to collect their own data, such as in black-box optimization and active learning, where their actions induce sequential feedback-loop shifts in the data distribution. Conformal prediction is a promising approach to uncertainty and risk quantification, but prior variants' validity guarantees have assumed some form of ``quasi-exchangeability'' on the data distribution, thereby excluding many types of sequential shifts. In this paper we prove that conformal prediction can theoretically be extended to *any* joint data distribution, not just exchangeable or quasi-exchangeable ones. Although the most general case is exceedingly impractical to compute, for concrete practical applications we outline a procedure for deriving specific conformal algorithms for any data distribution, and we use this procedure to derive tractable algorithms for a series of AI/ML-agent-induced covariate shifts. We evaluate the proposed algorithms empirically on synthetic black-box optimization and active learning tasks.

Despite advancements in robotic-assisted surgery, automating complex tasks like suturing remains challenging due to the need for adaptability and precision. Learning-based approaches, particularly reinforcement learning (RL) and imitation learning (IL), require realistic simulation environments for efficient data collection. However, current platforms often include only relatively simple, non-dexterous manipulations and lack the flexibility required for effective learning and generalization. We introduce SurgicAI, a novel platform for development and benchmarking that addresses these challenges by providing the flexibility to accommodate both modular subtasks and more importantly task decomposition in RL-based surgical robotics. Compatible with the da Vinci Surgical System, SurgicAI offers a standardized pipeline for collecting and utilizing expert demonstrations. It supports the deployment of multiple RL and IL approaches, and the training of both singular and compositional subtasks in suturing scenarios, featuring high dexterity and modularization. Meanwhile, SurgicAI sets clear metrics and benchmarks for the assessment of learned policies. We implemented and evaluated multiple RL and IL algorithms on SurgicAI. Our detailed benchmark analysis underscores SurgicAI's potential to advance policy learning in surgical robotics. Details: https://github.com/surgical-robotics-ai/SurgicAI

Appropriate reliance is critical to achieving synergistic human-AI collaboration. For instance, when users over-rely on AI assistance, their human-AI team performance is bounded by the model's capability. This work studies how the presentation of model uncertainty may steer users' decision-making toward fostering appropriate reliance. Our results demonstrate that showing the calibrated model uncertainty alone is inadequate. Rather, calibrating model uncertainty and presenting it in a frequency format allow users to adjust their reliance accordingly and help reduce the effect of confirmation bias on their decisions. Furthermore, the critical nature of our skin cancer screening task skews participants' judgment, causing their reliance to vary depending on their initial decision. Additionally, step-wise multiple regression analyses revealed how user demographics such as age and familiarity with probability and statistics influence human-AI collaborative decision-making. We discuss the potential for model uncertainty presentation, initial user decision, and user demographics to be incorporated in designing personalized AI aids for appropriate reliance.

Abstract Computational linguistics models commonly target the prediction of discrete—categorical—labels. When assessing how well-calibrated these model predictions are, popular evaluation schemes require practitioners to manually determine a binning scheme: grouping labels into bins to approximate true label posterior. The problem is that these metrics are sensitive to binning decisions. We consider two solutions to the binning problem that apply at the stage of data annotation: collecting either distributed (redundant) labels or direct scalar value assignment. In this paper, we show that although both approaches address the binning problem by evaluating instance-level calibration, direct scalar assignment is significantly more cost-effective. We provide theoretical analysis and empirical evidence to support our proposal for dataset creators to adopt scalar annotation protocols to enable a higher-quality assessment of model calibration.

Autonomous cars are indispensable when humans go further down the hands-free route. Although existing literature highlights that the acceptance of the autonomous car will increase if it drives in a human-like manner, sparse research offers the naturalistic experience from a passenger's seat perspective to examine the humanness of current autonomous cars. The present study tested whether the AI driver could create a human-like ride experience for passengers based on 69 participants' feedback in a real-road scenario. We designed a ride experience-based version of the non-verbal Turing test for automated driving. Participants rode in autonomous cars (driven by either human or AI drivers) as a passenger and judged whether the driver was human or AI. The AI driver failed to pass our test because passengers detected the AI driver above chance. In contrast, when the human driver drove the car, the passengers' judgement was around chance. We further investigated how human passengers ascribe humanness in our test. Based on Lewin's field theory, we advanced a computational model combining signal detection theory with pre-trained language models to predict passengers' humanness rating behaviour. We employed affective transition between pre-study baseline emotions and corresponding post-stage emotions as the signal strength of our model. Results showed that the passengers' ascription of humanness would increase with the greater affective transition. Our study suggested an important role of affective transition in passengers' ascription of humanness, which might become a future direction for autonomous driving.

Learning an optimal policy from offline data is notoriously challenging, which requires the evaluation of the learning policy using data pre-collected from a static logging policy. We study the policy optimization problem in offline contextual bandits using policy gradient methods. We employ a distributionally robust policy gradient method, DROPO, to account for the distributional shift between the static logging policy and the learning policy in policy gradient. Our approach conservatively estimates the conditional reward distributional and updates the policy accordingly. We show that our algorithm converges to a stationary point with rate $O(1/T)$, where $T$ is the number of time steps. We conduct experiments on real-world datasets under various scenarios of logging policies to compare our proposed algorithm with baseline methods in offline contextual bandits. We also propose a variant of our algorithm, DROPO-exp, to further improve the performance when a limited amount of online interaction is allowed. Our results demonstrate the effectiveness and robustness of the proposed algorithms, especially under heavily biased offline data.

We study the efficient estimation of predictive confidence intervals for black-box predictors when the common data exchangeability (e.g., i.i.d.) assumption is violated due to potentially feedback-induced shifts in the input data distribution. That is, we focus on standard and feedback covariate shift (FCS), where the latter allows for feedback dependencies between train and test data that occur in many decision-making scenarios like experimental design. Whereas prior conformal prediction methods for this problem are in general either extremely computationally demanding or make inefficient use of labeled data, we propose a collection of methods based on the jackknife+ that achieve a practical balance of computational and statistical efficiency. Theoretically, our proposed JAW-FCS method extends the rigorous, finite-sample coverage guarantee of the jackknife+ to FCS. We moreover propose two tunable relaxations to JAW-FCS's computation that maintain finite-sample guarantees: one using only $K$ leave-one-out models (JAW-$K$LOO) and a second building on $K$-fold cross validation+ (WCV+). Practically, we demonstrate that JAW-FCS and its computational relaxations outperform state-of-the-art baselines on a variety of real-world datasets under standard and feedback covariate shift, including for biomolecular design and active learning tasks.

We propose a framework for learning calibrated uncertainties under domain shifts, considering the case where the source (training) distribution differs from the target (test) distribution. We detect such domain shifts through the use of a differentiable density ratio estimator and train it together with the task network, composing an adjusted softmax predictive form that concerns the domain shift. In particular, the density ratio estimator yields a density ratio that reflects the closeness of a target (test) sample to the source (training) distribution. We employ it to adjust the uncertainty of prediction in the task network. This idea of using the density ratio is based on the distributionally robust learning (DRL) framework, which accounts for the domain shift through adversarial risk minimization. We demonstrate that our proposed method generates calibrated uncertainties that benefit many downstream tasks, such as unsupervised domain adaptation (UDA) and semi-supervised learning (SSL). On these tasks, methods like self-training and FixMatch use uncertainties to select confident pseudo-labels for re-training. Our experiments show that the introduction of DRL leads to significant improvements in cross-domain performance. We also demonstrate that the estimated density ratios show an agreement with the human selection frequencies, suggesting a positive correlation with a proxy of human perceived uncertainties.

We investigate model calibration in the setting of zero-shot cross-lingual transfer with largescale pre-trained language models.The level of model calibration is an important metric for evaluating the trustworthiness of predictive models.There exists an essential need for model calibration when natural language models are deployed in critical tasks.We study different post-training calibration methods in structured and unstructured prediction tasks.We find that models trained with data from the source language become less calibrated when applied to the target language and that calibration errors increase with intrinsic task difficulty and relative sparsity of training data.Moreover, we observe a potential connection between the level of calibration error and an earlier proposed measure of the distance from English to other languages.Finally, our comparison demonstrates that among other methods Temperature Scaling (TS) generalizes well to distant languages, but TS fails to calibrate more complex confidence estimation in structured predictions compared to more expressive alternatives like Gaussian Process Calibration.

Contemporary vision benchmarks predominantly consider tasks on which humans can achieve near-perfect performance. However, humans are frequently presented with visual data that they cannot classify with 100% certainty, and models trained on standard vision benchmarks achieve low performance when evaluated on this data. To address this issue, we introduce a procedure for creating datasets of ambiguous images and use it to produce SQUID-E ("Squidy"), a collection of noisy images extracted from videos. All images are annotated with ground truth values and a test set is annotated with human uncertainty judgments. We use this dataset to characterize human uncertainty in vision tasks and evaluate existing visual event classification models. Experimental results suggest that existing vision models are not sufficiently equipped to provide meaningful outputs for ambiguous images and that datasets of this nature can be used to assess and improve such models through model training and direct evaluation of model calibration. These findings motivate large-scale ambiguous dataset creation and further research focusing on noisy visual data.

We propose \textbf{JAWS}, a series of wrapper methods for distribution-free uncertainty quantification tasks under covariate shift, centered on the core method \textbf{JAW}, the \textbf{JA}ckknife+ \textbf{W}eighted with data-dependent likelihood-ratio weights. JAWS also includes computationally efficient \textbf{A}pproximations of JAW using higher-order influence functions: \textbf{JAWA}. Theoretically, we show that JAW relaxes the jackknife+'s assumption of data exchangeability to achieve the same finite-sample coverage guarantee even under covariate shift. JAWA further approaches the JAW guarantee in the limit of the sample size or the influence function order under common regularity assumptions. Moreover, we propose a general approach to repurposing predictive interval-generating methods and their guarantees to the reverse task: estimating the probability that a prediction is erroneous, based on user-specified error criteria such as a safe or acceptable tolerance threshold around the true label. We then propose \textbf{JAW-E} and \textbf{JAWA-E} as the repurposed proposed methods for this \textbf{E}rror assessment task. Practically, JAWS outperform state-of-the-art predictive inference baselines in a variety of biased real world data sets for interval-generation and error-assessment predictive uncertainty auditing tasks.

Making predictions that are fair with regard to protected attributes (race, gender, age, etc.) has become an important requirement for classification algorithms. Existing techniques derive a fair model from sampled labeled data relying on the assumption that training and testing data are identically and independently drawn (iid) from the same distribution. In practice, distribution shift can and does occur between training and testing datasets as the characteristics of individuals interacting with the machine learning system change. We investigate fairness under covariate shift, a relaxation of the iid assumption in which the inputs or covariates change while the conditional label distribution remains the same. We seek fair decisions under these assumptions on target data with unknown labels. We propose an approach that obtains the predictor that is robust to the worst-case testing performance while satisfying target fairness requirements and matching statistical properties of the source data. We demonstrate the benefits of our approach on benchmark prediction tasks.

We address the problem of active learning under label shift: when the class proportions of source and target domains differ. We introduce a "medial distribution" to incorporate a tradeoff between importance weighting and class-balanced sampling and propose their combined usage in active learning. Our method is known as Mediated Active Learning under Label Shift (MALLS). It balances the bias from class-balanced sampling and the variance from importance weighting. We prove sample complexity and generalization guarantees for MALLS which show active learning reduces asymptotic sample complexity even under arbitrary label shift. We empirically demonstrate MALLS scales to high-dimensional datasets and can reduce the sample complexity of active learning by 60% in deep active learning tasks.

Tracking and collecting fast-evolving online discussions provides vast data for studying social media usage and its role in people's public lives. However, collecting social media data using a static set of keywords fails to satisfy the growing need to monitor dynamic conversations and to study fast-changing topics. We propose a dynamic keyword search method to maximize the coverage of relevant information in fast-evolving online discussions. The method uses word embedding models to represent the semantic relations between keywords and predictive models to forecast the future trajectory of keywords. We also implement a visual user interface to aid in the decision making process in each round of keyword updates. This allows for both human-assisted tracking and fully-automated data collection. In simulations using historical #MeToo data in 2017, our human-assisted tracking method outperforms the traditional static baseline method significantly, achieving 37.1% improvement in F-1 score in the task of tracking the top trending keywords. We conduct a contemporary case study to cover dynamic conversations about the recent Presidential Inauguration and to test the dynamic data collection system. Our case studies reflect the effectiveness of our process and also points to the potential challenges in future deployment.

Active learning methods, like uncertainty sampling, combined with probabilistic prediction techniques have achieved success in various problems like image classification and text classification. For more complex multivariate prediction tasks, the relationships between labels play an important role in designing structured classifiers with better performance. However, computational time complexity limits prevalent probabilistic methods from effectively supporting active learning. Specifically, while non-probabilistic methods based on structured support vector ma-chines can be tractably applied to predicting cuts and bipartite matchings, conditional random fields are intractable for these structures. We propose an adversarial approach for active learning with structured prediction domains that is tractable for cuts and matching. We evaluate this approach algorithmically in two important structured prediction problems: multi-label classification and object tracking in videos. We demonstrate better accuracy and computational efficiency for our proposed method.

In many learning settings, the source data available to train a regression model differs from the target data it encounters when making predictions due to input distribution shift. Appropriately dealing with this situation remains an important challenge. Existing methods attempt to “reweight” the source data samples to better represent the target domain, but this introduces strong inductive biases that are highly extrapolative and can often err greatly in practice. We propose a robust approach for regression under covariate shift that embraces the uncertainty resulting from sample selection bias by producing regression models that are explicitly robust to it. We demonstrate the benefits of our approach on a number of regression tasks.

Recently proposed adversarial classification methods have shown promising results for cost sensitive and multivariate losses. In contrast with empirical risk minimization (ERM) methods, which use convex surrogate losses to approximate the desired non-convex target loss function, adversarial methods minimize non-convex losses by treating the properties of the training data as being uncertain and worst case within a minimax game. Despite this difference in formulation, we recast adversarial classification under zero-one loss as an ERM method with a novel prescribed loss function. We demonstrate a number of theoretical and practical advantages over the very closely related hinge loss ERM methods. This establishes adversarial classification under the zero-one loss as a method that fills the long standing gap in multiclass hinge loss classification, simultaneously guaranteeing Fisher consistency and universal consistency, while also providing dual parameter sparsity and high accuracy predictions in practice.

Existing approaches to active learning are generally optimistic about their certainty with respect to data shift between labeled and unlabeled data. They assume that unknown datapoint labels follow the inductive biases of the active learner. As a result, the most useful datapoint labels—ones that refute current inductive biases—are rarely solicited. We propose a shift-pessimistic approach to active learning that assumes the worst-case about the unknown conditional label distribution. This closely aligns model uncertainty with generalization error, enabling more useful label solicitation. We investigate the theoretical benefits of this approach and demonstrate its empirical advantages on probabilistic binary classification tasks.

To facilitate interaction with people, robots must not only recognize current actions, but also infer a person's intentions and future behavior. Recent advances in depth camera technology have significantly improved human motion tracking. However, the inherent high dimensionality of interacting with the physical world makes efficiently forecasting human intention and future behavior a challenging task. Predictive methods that estimate uncertainty are therefore critical for supporting appropriate robotic responses to the many ambiguities posed within the human-robot interaction setting. We address these two challenges, high dimensionality and uncertainty, by employing predictive inverse optimal control methods to estimate a probabilistic model of human motion trajectories. Our inverse optimal control formulation estimates quadratic cost functions that best rationalize observed trajectories framed as solutions to linear-quadratic regularization problems. The formulation calibrates its uncertainty from observed motion trajectories, and is efficient in high-dimensional state spaces with linear dynamics. We demonstrate its effectiveness on a task of anticipating the future trajectories, target locations and activity intentions of hand motions.