Alignment, interpretability, red-teaming, bias and privacy: the research on what these systems do when they misbehave.
arXiv:2605. 26937v2 Announce Type: replace-cross Abstract: Parametric knowledge in large language models (LLMs) is a cornerstone of their success, yet remains poorly understood.
arXiv:2608. 00143v2 Announce Type: replace-cross Abstract: Automated attack chain generation is critical for modern cybersecurity, yet manual construction fails to scale as adversary behaviors expand.
arXiv:2608. 01043v2 Announce Type: replace-cross Abstract: We report a counter-intuitive interaction between image inputs and existing black-box defenses on Vision--Language Models (VLMs): pairing an encoded jailbreak prompt with an unrelated decoy image can sharply lower attack success rate (ASR).
arXiv:2608. 07725v1 Announce Type: new Abstract: Average-reward reinforcement-learning regret is known up to logarithmic factors, but the numerical content of published guarantees is difficult to compare because probability mode, structural parameter, logarithmic normalization, prior information, and planning assumptions differ.
arXiv:2608. 07935v1 Announce Type: new Abstract: On-policy self-distillation (OPSD) adapts a language model by distilling guidance from a frozen teacher on trajectories sampled from the student.
arXiv:2608. 08008v1 Announce Type: new Abstract: Process reward models (PRMs) score intermediate reasoning steps and are widely used for search, ranking, and training, but optimization can exploit these learned proxies by increasing reward while turning correct reasoning into incorrect reasoning.
arXiv:2608. 08224v1 Announce Type: new Abstract: Reinforcement learning post-training unlocks complex reasoning in LLMs.
arXiv:2608. 08202v1 Announce Type: new Abstract: Data-centric curation pipelines frequently rely on model confidence scores to flag and filter noisy or mislabeled training instances.
arXiv:2608. 08350v1 Announce Type: new Abstract: The initialisation of deep neural networks determines whether information and gradients can propagate across depth, yet a unified theory connecting these properties to learning dynamics remains elusive.
arXiv:2608. 08536v1 Announce Type: new Abstract: Circuit localization is a mechanistic interpretability task whose goal is to identify a sparse subgraph of a transformer's computation graph sufficient to reproduce a particular behavior.
arXiv:2608. 08542v1 Announce Type: new Abstract: Model merging has become the default way to give an aligned language model new skills without retraining: a practitioner folds task vectors from math, code, or domain specialists into a safety-aligned base using task arithmetic, TIES, or DARE.
arXiv:2608. 08624v1 Announce Type: new Abstract: Domain generalization (DG) and neural network pruning are conventionally treated as distinct objectives, targeting out-of-distribution (OOD) robustness and model efficiency, respectively.
arXiv:2608. 08920v1 Announce Type: new Abstract: Perioperative risk prediction models are often limited by narrow surgical populations, incomplete intraoperative data, poor calibration, and limited interpretability.
arXiv:2608. 09082v1 Announce Type: new Abstract: Spatiotemporal prediction on graph-structured data is central to traffic forecasting and environmental monitoring, yet decentralized and heterogeneous data complicate both sequence modeling and collaborative training.
arXiv:2608. 09366v1 Announce Type: new Abstract: Large-scale learning systems often face the challenge of balancing multiple, potentially competing objectives, such as fairness, accuracy, and latency.
arXiv:2608. 09830v1 Announce Type: new Abstract: Body-focused repetitive behaviors, such as hair pulling and skin picking, are compulsive motor actions commonly associated with obsessive-compulsive and anxiety disorders.
arXiv:2608. 07522v1 Announce Type: cross Abstract: We present a structured review of commonly used Explainable machine learning (XML) methodologies, including global and local interpretability tools such as SHapley Additive exPlanations (SHAP), Local Interpretable Model-Agnostic Explanations (LIME), Partial Dependence Plots (PDP), and Individual Conditional Expectation (ICE) plots.
arXiv:2608. 07583v1 Announce Type: cross Abstract: Multi-agent LLM systems route among model-backed advisors, yet a deployer rarely knows before shipping whether routing will help at all.
arXiv:2608. 07586v1 Announce Type: cross Abstract: Semi-supervised Class Incremental Learning (SSCIL) is a severe challenge for neural networks, and it is hardest in the exemplar-free setting where no past data may be stored.
arXiv:2608. 07643v1 Announce Type: cross Abstract: Traffic data collection is dominated today by deep object detectors followed by tracking-by-detection, a pipeline that presupposes what is often missing in practice: a detector already trained on the class one wants to count.