2315 papers
This paper introduces a realistic probabilistic framework based on the "(k, )-unstable" assumption to derive data-informed safety certificates for SmoothLLM, overcoming the limitations of strict theoretical guarantees and providing actionable defense guarantees against diverse jailbreaking attacks.
This paper introduces an automated pipeline for generating large-scale, psychologically-grounded multi-turn jailbreak datasets using Foot-in-the-Door techniques, revealing that while models like GPT are significantly vulnerable to conversational history, others like Gemini 2.5 Flash demonstrate exceptional resilience against such narrative-based manipulation.
This paper proposes "Shortcut Invariance," a targeted Jacobian regularization method that improves out-of-distribution generalization by injecting anisotropic noise into a disentangled latent space to flatten decision boundaries along shortcut-aligned axes, thereby eliminating the need for explicit shortcut labels or conflicting training samples.
This paper demonstrates that a $50,000 Kaggle competition successfully crowdsourced diverse machine learning architectures for subgrid parameterization, which, when coupled with a full-physics climate model, achieved reproducible online stability and state-of-the-art performance, marking a significant milestone in advancing hybrid physics-ML climate simulations.
The paper introduces CRAwDAD, a dual-agent debate framework that enhances causal inference in reasoning language models by facilitating structured dialogue and adversarial critique between agents, significantly improving accuracy on the CLadder benchmark across all levels of Pearl's causal ladder.
This study introduces ForamDeepSlice, a high-accuracy deep learning framework that combines an ensemble of ConvNeXt-Large and EfficientNetV2-Small models with a rigorous specimen-level split dataset to achieve 95.64% accuracy in classifying foraminifera species from 2D micro-CT slices, while also providing an interactive dashboard for real-time identification and 3D matching.
The paper introduces AltNet, a twin-network reinforcement learning framework that restores plasticity through periodic parameter resets without causing performance drops, thereby achieving higher sample efficiency and stability in high-dimensional control tasks compared to existing reset-based methods.
The paper introduces MSPT, a novel architecture that leverages ball trees for efficient spatial partitioning and a dual-scale attention mechanism to enable scalable, memory-efficient, and high-accuracy physical simulations of millions of points on a single GPU across diverse industrial applications.
This paper proposes Dual Randomized Smoothing, a novel framework that overcomes the limitations of global noise variance by introducing input-dependent noise variances via a locally constant variance estimator, thereby achieving superior certified robustness across both small and large perturbation radii on CIFAR-10 and ImageNet.
This paper introduces DROCO, a novel dual-robust cross-domain offline reinforcement learning algorithm that addresses both train-time and test-time dynamics shifts by employing a robust cross-domain Bellman operator alongside dynamic value penalty and Huber loss to enhance policy robustness and prevent value estimation errors.
The paper introduces GoRL, an algorithm-agnostic framework that stabilizes online reinforcement learning with expressive generative policies by decoupling optimization in a tractable latent space from action synthesis via a conditional generative decoder, achieving superior performance on challenging continuous-control tasks.
This paper introduces Sparse Isotonic Shapley Regression (SISR), a unified framework that simultaneously learns a monotonic transformation to restore additivity and enforces sparsity constraints to provide robust, efficient, and theoretically grounded feature attributions for nonlinear, high-dimensional Explainable AI.
This paper presents a two-step generative data augmentation framework combining rule-based mask warping and unpaired image-to-image translation to address the scarcity of masked face datasets, achieving performance improvements with minimal training data while explicitly noting its origins as a resource-constrained coursework project that lacked downstream quantitative evaluation.
The paper introduces SALVE, a unified framework that combines sparse autoencoders and feature-level saliency mapping to discover, validate, and precisely edit neural network weights, enabling interpretable and robust control over both convolutional and transformer-based models.
The paper introduces LaMer, a Meta-RL framework that enhances language agents' ability to actively explore and adapt to novel environments at test time through cross-episode training and in-context policy reflection, significantly outperforming standard RL baselines across diverse tasks.
Re-Depth Anything is a test-time self-supervised framework that enhances monocular depth estimation by fusing foundation models with large-scale 2D diffusion priors to perform label-free refinement via generative re-lighting and Score Distillation Sampling, achieving state-of-the-art results without direct depth tensor optimization.
This paper proposes three complementary strategies—pre-training tailored to Newton-Kaczmarz updates, training on disjoint data subsets with model merging, and FPGA-based parallelization—to overcome the sequential bottlenecks in Kolmogorov-Arnold network training and significantly accelerate convergence.
The paper proposes "Latent Sculpting," a hierarchical two-stage architecture that combines a Transformer-based encoder with a Binary Latent Sculpting loss and a Masked Autoregressive Flow to enforce explicit geometric boundaries on benign data, achieving robust zero-shot generalization and high detection rates for out-of-distribution cyberattacks on the CIC-IDS-2017 benchmark.
This paper proposes FedORA, a primal-dual optimization framework that enables efficient and theoretically certified sample and label unlearning in Vertical Federated Learning by introducing a novel uncertainty-promoting loss function and adaptive strategies to minimize computational overhead while preserving model utility.
This paper proposes a process mining-based method that analyzes online gaming network traffic to unsupervisedly characterize device states as interpretable Petri nets and accurately classify different video games, demonstrating its effectiveness through the UPSIDE case study involving Clash Royale and Rocket League.