arXiv paper on ARES, scalable gradient inversion attack in federated learning via activation recovery, demonstrating privacy risks in FL.
arXiv paper proposing benchmarking framework for RL algorithms using stochastic converse optimality to generate systems with known optimal policies.
arXiv paper introducing DSS-GAN, first GAN using Mamba backbone for class-conditional image synthesis with novel Directional Latent Routing mechanism.
arXiv paper on flow matching policies with entropy regularization for diffusion-based reinforcement learning, improving policy gradient computation.
arXiv paper on identifying undervalued football players using market dynamics data and NLP-derived news signals to detect objective mispricing.
arXiv paper on LLM trading agents with anonymization framework to detect memorization bias and validate genuine market understanding vs. ticker recall.
arXiv paper benchmarking 256 LLM-based embedding pipeline configurations for tabular prediction, evaluating preprocessing strategies, embedding models, and downstream models.
arXiv paper studying attention sinks in Transformers from backpropagation perspective, showing attention sinks induce gradient concentration under causal masking.
RangeAD leverages primary model's learned representations for efficient on-model anomaly detection without separate AD model.
Analysis of dropout-induced variability in transformer models via Monte Carlo sampling to assess uncertainty awareness and reliability.
FedDistRL formalizes federated distributional reinforcement learning with quantile value functions for safety-critical applications.
ULCMOD framework discovers and disentangles functional modules in LLMs through unsupervised cross-layer analysis for interpretability.
SymPINN framework embeds group-theory symmetries into physics-informed neural networks for tensegrity structure dynamics simulation.
DiscoGen procedural generator creates diverse algorithm discovery tasks to improve evaluation of AutoML systems and algorithm design optimization.
RAMP uses reinforcement learning to assign per-layer bit widths for mixed-precision quantization of LLMs on resource-constrained devices.
Weight clustering technique for LLMs showing relative rank of weights matters more than precise magnitudes for model compression and efficiency.
CARE method converts grouped-query attention to multi-head latent attention for efficient LLM inference using covariance-aware rank decomposition.
Framework for rapid adaptation in reinforcement learning where policy and value functions share low-dimensional embeddings for novel task generalization.
MUD (MomentUm Decorrelation) optimizer for faster transformer training using whitening approach as alternative to polar decomposition methods like Muon.
Multi-agent reinforcement learning framework for radiology report generation with clinically verifiable rewards.
Physics-informed graph attention network for real-time AC power flow prediction with continual learning.
Foundation model for converting EEG signals to clinical text interpretations via spectro-spatial grounding.
Controlled comparison of 9 deep learning architectures for financial time-series forecasting across 918 experiments.
EEG-based brain interface enables LLM interaction for users with speech/motor impairments via neural signal decoding.
Study demonstrating that LLM agents can autonomously infer CoT monitoring from blocking feedback, creating risks for evasion of reasoning oversight.
Method for extracting and clustering traffic scenarios from real-world highway data using conditional VAE for autonomous vehicle testing.
DeepStage: Deep reinforcement learning framework for autonomous defense against multi-stage APT attacks using provenance graphs and stage estimation.
Quantum transfer learning architecture combining pretrained classical models with variational quantum classifiers for image classification on noisy hardware.
TorchNWP: Compiler library tool for coupling AI models with traditional numerical models, enabling Fortran-Python interoperability for weather prediction.
Reward prediction model for robot manipulation using vision foundation models to infer dense task rewards from camera images without privileged state.
Evaluation metric for generative models assessing whether synthetic data preserves multivariate dependence structures for downstream inference tasks.
DesertFormer: Transformer-based semantic segmentation pipeline for off-road desert terrain classification in autonomous navigation systems.
Multi-fidelity surrogate modeling framework for airfoil optimization combining low-fidelity simulations with Gaussian processes and genetic algorithms.
Ensemble self-training approach for unsupervised neural machine translation using multiple models with auxiliary languages and token-level ensemble decoding.
Auto-Prov: End-to-end framework using LLMs to construct provenance graphs from system logs for anomaly detection and threat interpretation.
Framework for locating knowledge in mixture-of-experts LLMs by analyzing cross-lingual inconsistencies, advancing interpretability of expert routing.
Self-supervised learning method for medical image segmentation using contrastive learning and counterfactual generation to handle imperfect AI labels.
Multi-agent routing architecture for AI reasoning systems with dynamic execution graphs, addressing cascade failure propagation in agent delegation networks.
Safe reinforcement learning framework for robots using temporal logic constraints to enforce safety and operational requirements during training.
TAP-GPT uses pretrained LLMs for few-shot Alzheimer's disease prediction from multimodal biomedical tabular data.
OPERA framework for data pruning to improve efficiency and effectiveness of dense retriever finetuning.
Adaptive contracts framework for cost-effective AI delegation, balancing evaluation noise against evaluation costs.
LLM-driven pipeline for anonymizing text by replacing PII with realistic surrogates while preserving data utility.
Approach for developing Tharu language LLM using synthetic data generation and human validation to address low-resource language gap.
Analysis of multimodal LLM segmentation capabilities through layerwise probing and attention mechanisms.
Red-teaming alignment framework (CRAFT) that improves LLM robustness against jailbreaks by optimizing hidden representations.
Multi-agent RL framework for dynamic memory controller optimization with explainable energy and latency objectives.
Offline RL framework (PIER) for fuel-efficient maritime routing using physics-informed models and historical vessel data.
Multimodal LLM framework for ride-hailing dispute resolution combining visual and logical reasoning with transparency.
AI coding agent that bootstraps itself by re-implementing its own specification, demonstrating meta-circular properties similar to compiler bootstrapping.