robustness

Flow-JEPA introduces a conditional flow matching approach to JEPA world models, improving robustness and accuracy in predicting future latent states under noisy conditions.

SnapBench introduces a paired corruption benchmark for robust snap-and-ask multimodal retrieval on mobile, revealing image noise as a key degradation factor and proposing an adaptive fusion method for modality reliability.

This paper proposes a novel adversarial training method that eliminates input gradients by using low-rank Householder expansions, reducing computational cost while achieving robustness comparable to standard techniques for small perturbations.

The paper audits large language models on their refusal and fabrication behavior in clinical pain speech transcripts, finding that authority-framed prompts lead to confident fabrication in models like Gemini 2.5 Flash and Llama 3.1 8B, while cooperative prompting shows robust abstention.

The article discusses how agentic coding and AI could transform open-source software by making coding less scarce, using biological analogies to contrast short-term efficiency with long-term robustness.

This paper evaluates self-supervised pretraining for tabular models under label scarcity and missing data, finding mixed efficacy but reliable improvements under test-time missingness.

This paper introduces a methodological framework for auditing the robustness and fidelity of post-hoc explainable AI tools like SHAP and LIME, combining these metrics into a Trust Score. It applies the framework to a food security dataset in Madagascar, highlighting the necessity of auditing XAI outputs for trustworthy decision-making in sensitive domains.

This paper empirically studies how lexical perturbations disrupt large language model reasoning through attention diversion, finding that character-level noise significantly degrades performance while filler insertions have little effect.

This paper introduces AnTrap, a benchmark for evaluating the robustness of Android GUI agents against runtime anomalies, revealing universal vulnerabilities and differentiating between learnable traps and intrinsic reasoning limitations.

This paper proposes Jacobian-guided noise injection, a training strategy that improves quantization robustness in large language models by injecting noise into pre-attention logits with variance derived from the Jacobian norm, leading to significant performance gains in low-bit quantization settings.

This paper introduces a locally tokenized generative model for robust watermarking in multivariate time-series data, addressing reliability issues under post-editing attacks by using bounded temporal neighborhoods for token recovery.

This paper proposes an iterative proxy correction framework to enhance robustness in multimodal sentiment analysis when dealing with incomplete or corrupted inputs by refining a language proxy for better sentiment prediction.

The paper introduces a failure-aware adversarial retrieval-augmented framework using contextual bandits to improve robustness in natural language understanding, with significant improvements on benchmarks like SNLI, ANLI, and MultiNLI.

MissDiag introduces a diagnostic evaluation framework that decomposes robustness in KGQA and KG-RAG systems under incomplete knowledge into typed evidence interventions for more interpretable comparisons.

This paper evaluates the robustness of AI code agents when codebases are perturbed with semantics-preserving transformations, revealing a jagged frontier where model performance varies unpredictably across different scaffolds and benchmarks.

RoBell-RVFL is a robust randomized neural network that uses a generalized bell membership function to handle class imbalance and noisy data, achieving superior performance on benchmark datasets.

This research paper investigates how data perturbations impact the accuracy and robustness of model cascades for efficient AI inference, identifying key failure modes and stressing the need for evaluation under distribution shift.

This paper introduces stochastic control policies (FS-TPS and LaS-TPS) for molecular transition path sampling, improving robustness and performance across biomolecular systems.

Introduces Action-Conditioned Predictive Consistency (ACPC), a diagnostic for JEPA world models that measures how clean and perturbed observations diverge under action-conditioned rollouts, with theoretical bounds on prediction error and planner cost. Experiments on visual control tasks validate the diagnostic across models like LeWM and PLDM.

Introduces the Wiggle Framework, a unified stress test for the epistemic stability of LLM judges, measuring mechanical consistency, single-turn conviction, and multi-turn persistence across 14 tasks. Finds that all 9 studied frontier models flip verdicts substantially under pressure, and that successful pressure is usually net-corrupting relative to ground truth.