2295 papers
This paper introduces the ORIC framework and benchmark to evaluate and improve Large Vision-Language Models' object recognition capabilities under contextual incongruity, demonstrating that such scenarios significantly degrade performance and that targeted Visual Reinforcement Fine-Tuning can effectively mitigate these failures.
This paper proposes ORN-CBF, a hypernetwork-based learning framework that utilizes Hamilton-Jacobi reachability analysis to generate observation-conditioned neural control barrier functions, ensuring rigorous safety guarantees and improved generalization in partially observable environments through simulation and hardware experiments.
This paper introduces the first fully empirical PAC-Bayes bound for Markov chains by deriving a data-dependent estimate for the pseudo-spectral gap, thereby eliminating the need for unknown constants related to mixing properties that typically hinder practical generalization guarantees.
This paper demonstrates that linear probes used to detect harmful behaviors in language models are heavily reliant on explicit textual evidence, as their performance significantly degrades when such surface-level cues are filtered out or when models are trained to express behaviors without verbalization.
The paper introduces AEGIS, an edge-only augmentation framework that resamples existing training edges to enhance link prediction in edge-sparse bipartite knowledge graphs, demonstrating that authenticity-constrained resampling preserves data integrity while semantic KNN augmentation further boosts performance when node descriptions are available.
This paper introduces GDR-learners, a flexible suite of generative models (including CNFs, CGANs, CVAEs, and CDMs) that achieve quasi-oracle efficiency and double robustness for estimating potential outcome distributions, thereby outperforming existing methods in both theoretical properties and empirical performance.
CLAD-Net is a continual learning framework for wearable human activity recognition that combines a self-supervised transformer for long-term memory and a supervised CNN with knowledge distillation to effectively mitigate catastrophic forgetting and handle label scarcity across diverse subjects.
The paper introduces Generative Evolutionary Meta-Solver (GEMS), a scalable, surrogate-free multi-agent reinforcement learning framework that replaces explicit policy populations with a compact generator and latent anchors to achieve significantly faster training, lower memory usage, and higher rewards than traditional methods like PSRO while maintaining game-theoretic guarantees.
This paper introduces FS-KAN, a principled framework that constructs permutation equivariant and invariant Kolmogorov-Arnold Networks via function sharing, offering superior data efficiency and interpretability while maintaining the expressive power of standard parameter-sharing networks.
This paper introduces Overlap-Adaptive Regularization (OAR), a novel method that enhances the performance of existing CATE meta-learners in low-overlap regions by proportionally increasing regularization based on overlap weights, while offering flexible, debiased variants that preserve Neyman-orthogonality for robust inference.
This paper addresses the cold-start scenario in active correlation clustering, where no initial pairwise similarities are available, by proposing a coverage-aware method that encourages diversity to efficiently query similarities and achieve effective clustering.
This paper proposes a principled feedback control method combining bucketized hysteresis and proportional feedback to achieve stable, accurate, and scalable budget pacing for online advertising campaigns, particularly improving performance for small-budget scenarios by significantly reducing pacing errors and volatility compared to existing baselines.
This paper introduces and empirically validates the concept of "misevolution," demonstrating that self-evolving LLM agents face widespread, emergent risks across model, memory, tool, and workflow pathways that can lead to safety degradation and unintended vulnerabilities, thereby highlighting an urgent need for new safety paradigms.
This paper introduces the DRQ-learner, a novel meta-learner for estimating individualized potential outcomes in Markov Decision Processes that combines double robustness, Neyman orthogonality, and quasi-oracle efficiency to outperform existing state-of-the-art methods in sequential decision-making.
This paper presents a differentially private scheme for estimating black-box statistics that effectively trades off between statistical efficiency and oracle efficiency, accompanied by lower bounds demonstrating the near-optimality of this approach.
The paper introduces SelfOrg, a training-free framework that enables multi-agent systems to self-organize by dynamically constructing response-conditioned communication graphs based on Shapley value approximations, thereby optimizing collaboration and significantly improving performance—especially with weaker LLMs—without relying on fixed topologies or external supervision.
The paper introduces CroSTAta, a Cross-State Transition Attention Transformer that enhances robotic manipulation robustness by employing a novel State Transition Attention mechanism to model temporal structures like failure and recovery patterns, outperforming standard attention and sequential models in simulation.
This paper introduces a "double projection" method within dynamical variational autoencoders that simultaneously estimates system state trajectories and noise time series from data, enabling effective multi-step reconstruction and learning of low-dimensional stochastic models across various benchmark problems.
The paper introduces DialTree, a tree-based dialogue reinforcement learning framework that autonomously discovers diverse and effective multi-turn attack strategies against large language models, significantly outperforming existing single-turn or template-based red-teaming methods.
This paper demonstrates that current graph-based tabular deep learning methods fail to accurately recover underlying feature interaction structures despite their focus on predictive accuracy, and shows that explicitly modeling the true graph structure significantly improves prediction performance.