Xiao Wang

Stephan Rabanser, Ali Shahin Shamsabadi, Olive Franzese et al.

Cautious predictions -- where a machine learning model abstains when uncertain -- are crucial for limiting harmful errors in safety-critical applications. In this work, we identify a novel threat: a dishonest institution can exploit these mechanisms to discriminate or unjustly deny services under the guise of uncertainty. We demonstrate the practicality of this threat by introducing an uncertainty-inducing attack called Mirage, which deliberately reduces confidence in targeted input regions, thereby covertly disadvantaging specific individuals. At the same time, Mirage maintains high predictive performance across all data points. To counter this threat, we propose Confidential Guardian, a framework that analyzes calibration metrics on a reference dataset to detect artificially suppressed confidence. Additionally, it employs zero-knowledge proofs of verified inference to ensure that reported confidence scores genuinely originate from the deployed model. This prevents the provider from fabricating arbitrary model confidence values while protecting the model's proprietary details. Our results confirm that Confidential Guardian effectively prevents the misuse of cautious predictions, providing verifiable assurances that abstention reflects genuine model uncertainty rather than malicious intent.

Sheng Jia, Xiao Wang, Shiva Prasad Kasiviswanathan

Although LLMs have demonstrated improved performance by scaling parallel test-time compute, doing so relies on generating reasoning paths that are both diverse and accurate. For challenging problems, the forking tokens that trigger diverse yet correct reasoning modes are typically deep in the sampling tree. Consequently, common strategies to encourage diversity, such as temperature scaling, encounter a worsened trade-off between diversity and accuracy. Motivated by this challenge, we treat parallel reasoning as a set-of-next-token-prediction problem, and incorporate a set-based global loss into Supervised Fine-Tuning (SFT) using self-supervised bipartite matching between our global forking tokens and unique reasoning traces. We observe that, while naive fine-tuning with multiple reasoning traces collapses these unique reasoning modes, our proposed method, Set Supervised Fine-Tuning (SSFT), preserves these modes and produces emergent global forking tokens. Experiments on multiple reasoning benchmarks show that our SSFT consistently outperforms SFT under both Pass@1 and Cons@k metrics.

Jennie Rogers, Elizabeth Adetoro, Johes Bater et al.

Electronic health records represent a rich and growing source of clinical data for research. Privacy, regulatory, and institutional concerns limit the speed and ease of sharing this data. VaultDB is a framework for securely computing SQL queries over private data from two or more sources. It evaluates queries using secure multiparty computation: cryptographic protocols that evaluate a function such that the only information revealed from running it is the query answer. We describe the development of a HIPAA-compliant version of VaultDB on the Chicago Area Patient Centered Outcomes Research Network (CAPriCORN). This multi-institutional clinical research network spans the electronic health records of nearly 13M patients over hundreds of clinics and hospitals in the Chicago metropolitan area. Our results from deploying at three health systems within this network show its efficiency and scalability for distributed clinical research analyses without moving patient records from their site of origin.

Lixu Wang, Shichao Xu, Xiao Wang et al.

Federated learning (FL) is a promising approach for training decentralized data located on local client devices while improving efficiency and privacy. However, the distribution and quantity of the training data on the clients' side may lead to significant challenges such as class imbalance and non-IID (non-independent and identically distributed) data, which could greatly impact the performance of the common model. While much effort has been devoted to helping FL models converge when encountering non-IID data, the imbalance issue has not been sufficiently addressed. In particular, as FL training is executed by exchanging gradients in an encrypted form, the training data is not completely observable to either clients or servers, and previous methods for class imbalance do not perform well for FL. Therefore, it is crucial to design new methods for detecting class imbalance in FL and mitigating its impact. In this work, we propose a monitoring scheme that can infer the composition of training data for each FL round, and design a new loss function -- \textbf{Ratio Loss} to mitigate the impact of the imbalance. Our experiments demonstrate the importance of acknowledging class imbalance and taking measures as early as possible in FL training, and the effectiveness of our method in mitigating the impact. Our method is shown to significantly outperform previous methods, while maintaining client privacy.