robustness

This survey explores how combinatorial optimization can be used to enhance trustworthy machine learning, covering topics like interpretability, robustness, fairness, and privacy with formal guarantees.

Introduces LiST, a training paradigm that uses Lipschitz constraints to achieve robust and calibrated neural networks, selecting optimal operating points on the accuracy-robustness Pareto front. Demonstrates competitive performance on CIFAR and Tiny-ImageNet.

This paper proposes a unified optimization framework to explain misclassifications and assess classifier robustness by sparse, interpretable instance alterations and a Tolerance Region Confusion Matrix.

This paper introduces a formal framework for distributed secret storage in networks, balancing survivability under failures and resistance to adversarial compromise, and derives exact survivability in terms of minimal information-carrying subgraphs.

This paper presents an empirical study comparing how different neural architectures (MLPs, CNNs, RNNs, pretrained transformers) degrade under temporal distribution shift across image and text domains, finding that models exploiting localized features degrade fastest while pretrained encoders drift more gradually.

Proposes exogenous dropout, a simple training intervention that randomly zeros entire exogenous channels, to improve robustness of time series forecasting models against corrupted covariates. The method matches or exceeds more complex architectures across multiple corruption types and domains.

This paper introduces Evidential Adversarial Training (EV-AT), a method that improves the robustness-uncertainty trade-off in classifiers by combining an evidence-based loss with robust evidence alignment, achieving state-of-the-art results on selective classification benchmarks.

This paper introduces the first systematic robustness benchmark for plasma diagnostic machine learning models using the TokaMark dataset, evaluating architectures across six sensor failure scenarios and introducing a Robustness Score for cross-architecture comparison.

This paper proposes a distributionally robust listwise preference optimization method for LLM alignment that handles ranking-label uncertainty, with a tractable objective and strong convergence guarantees.

FLYNN is a recurrent neural network derived from the fruit fly brain connectome for vision-based robot navigation. It achieves comparable performance to handcrafted networks while showing superior robustness to out-of-distribution data and sensory loss.

PRA-RAG is a provably robust aggregation algorithm for Retrieval-Augmented Generation that defends against poisoning attacks on retrieved texts. It uses geometric structures in the embedding space to identify robust subsets and provides theoretical bounds on attack impact, reducing attack success rate to as low as 1% while maintaining accuracy.

This paper explores deep reinforcement learning for attitude control of spacecraft during hypersonic re-entry. It demonstrates that state-of-the-art RL and hybrid controllers can outperform traditional PID controllers with gain scheduling, especially when dynamics randomization is used to improve robustness and generalization.

This paper introduces 'performative compliance' in LLMs, where models appear fair only when demographic identity is explicitly labeled but become less fair when identity must be inferred. The authors propose a cue-variation methodology and a Cue Visibility Gap metric to measure genuine versus superficial moral safety.

This paper presents Dual-Embedding Watermarking (DEW), a semantic watermarking scheme for LLMs that improves robustness against paraphrasing and translation by leveraging contextual and token-level embeddings. Experimental results show improved detection after paraphrasing and translation compared to prior methods.

This paper studies a lightweight prompting-based recovery approach for LLM dialogue agents when backend database calls fail, showing that the Guided-Retry strategy reduces hallucination by 50% on MultiWOZ and 42% on SGD across six model families.

Proposes Triospect, a three-dimensional framework that enhances AI-generated text detection robustness against 17 types of attacks, achieving 22.3% AUROC improvement over baselines.

This paper introduces TrajRS, an extension of Randomized Smoothing that provides certified robust radii for pedestrian trajectory predictors, offering verifiable safety guarantees against adversarial perturbations.

This paper proposes a symbolic feedback-driven iterative self-refinement framework to improve the robustness and reliability of large language models in long-horizon planning tasks. The method uses natural language prompting, a symbolic verifier, and a plan recognizer to enhance feasibility and correctness.

Aloe-Vision introduces a family of open medical Vision-Language Models trained on a quality-filtered mixture of medical and general data, along with a new benchmark CareQA-Vision for reliable evaluation. The models demonstrate competitive performance while highlighting vulnerabilities to adversarial inputs.

FoggyTrust is a hierarchical extension of FLTrust that localizes trust computation to fog nodes, improving robustness against Byzantine attacks in heterogeneous federated learning settings, achieving over 50% improvement on challenging attacks like Krum and Trim on CIFAR-10.