PRISM framework combines channel prioritization and semi-supervised domain adaptation for cross-subject EEG emotion recognition.
Self-supervised continual graph learning method using structure-aware optimal transport for sequential graph tasks.
GenDa addresses non-stationary skill semantics and generalization in unsupervised RL for skill-conditioned policy pre-training.
Proposes temporal fidelity framework for health signal forecasting that preserves oscillatory and phase dynamics.
Introduces spectral effective-rank entropy as diagnostic for critic complexity in actor-critic RL methods.
Benchmark for evaluating safety risks in AI-generated molecules, addressing toxicity and hazard detection.
Comparison of seven categorical encoding methods for high-cardinality fraud detection on IEEE-CIS benchmark dataset.
Linear transformers enable efficient in-context learning with reduced computational complexity while maintaining task generalization.
Diffusion models fine-tuned via RL for personalized recommender systems using interactive user feedback.
arXiv paper introducing prototype language models that make training data influence explicit and traceable for interpretability.
arXiv paper connecting structural equation modeling with machine learning for robustness analysis in survey research.
arXiv paper combining adaptive imitation learning and RL for improving LLM reasoning in molecular optimization.
arXiv paper on RL post-training for few-step flow-map generators using anchored stochastic composition.
arXiv paper on group-equivariant convolutional networks in hyperbolic space for visual representations.
arXiv paper on decision-focused learning for sparse portfolio optimization using end-to-end framework.
arXiv paper measuring singular structure in neural networks through directional-Fisher rate analysis.
arXiv paper on loss smoothing technique to stabilize neural network adaptation under distribution shift.
arXiv paper on multi-label node classification using label influence propagation on graphs.
arXiv paper on distributed online submodular maximization with bandit feedback across multiple agents.
arXiv paper using AdaBoosting for adapting text prompts in vision-language models with few labeled examples.
arXiv paper on generative models for low-budget black-box optimization with noisy or expensive evaluations.
arXiv paper combining active learning with unsupervised methods to detect anomalies in time series data.
arXiv paper using LLMs to guide ODE discovery and parameter inference from limited clinical data for rare diseases.
arXiv paper on MosaicKV, a dynamic 2D KV cache compression technique for serving long-context LLMs efficiently.
arXiv paper parallelizing discrete diffusion model sampling via τ-leaping algorithm in CTMC framework for faster generation.
arXiv paper on metrics for ERP-based brain-computer interfaces focusing on spelling rate accuracy.
arXiv paper proposing task-relevant representation decoupling for visual RL to improve generalization across environments.
arXiv paper on RL pre-training from videos using local motion patterns for better transfer across different agent morphologies.
Unsupervised pre-training method for RL using temporal correlations from large-scale video datasets to improve sample efficiency.
ML regression models for predicting carbon and nitrogen content in soil using near-infrared spectroscopy data.
Bayesian optimization method using multiple information sources for constrained design space exploration with small feasible regions.
arXiv paper on mixed-precision LLM quantization identifying alignment-diversity tradeoff and perplexity illusion phenomenon.
arXiv paper on Value Diffusion World Models combining diffusion models with model predictive control for latent planning.
arXiv paper proposing Generative Meta-Learning with Human Feedback (GMHF) framework for domain generalization.
arXiv paper on explainable AI for cancer drug response prediction using multivariate feature attribution methods.
arXiv paper on diffeomorphic optimization for solving objectives on low-dimensional manifolds using diffusion models.
arXiv paper on LeNEPA self-supervised learning method for time-series representation without data augmentation.
Study on automatic stress detection from speech in controlled test setting using machine learning.
arXiv paper on particle filtering for data assimilation using generative models as proposal distributions.
Seahorse: unified benchmarking framework for spatiotemporal point process models with standardized evaluation methodology.
arXiv paper showing model organism interpretability depends strongly on training methodology, affecting white-box interpretability evaluation.
arXiv paper proposing GSRQ for sub-1-bit KV cache quantization in large language models via gain-shape residual quantization.
arXiv paper on Gaussian process bandit optimization using quantum kernels for NISQ-era quantum tasks.
arXiv paper on AlphaEarth for spatio-temporal point-process forecasting using spatial context embeddings.
arXiv paper studying staleness effects in asynchronous RLHF systems with scaling laws for policy optimization.
arXiv paper on CausalMix for optimizing data mixture weights in LLM training using causal inference without retraining.
arXiv paper on SynLaD latent diffusion framework for generating synthesizable drug molecules with pharmacophore constraints.
arXiv paper proposing Muon optimizer as implicit residual connection during neural network training, explaining its effectiveness.
arXiv paper presenting ZO-Act for efficient zeroth-order fine-tuning of large language models using activation-informed low-rank subspaces.
arXiv paper on GAIA, a geometry-adaptive neural operator learning method for PDEs including boundary value and inverse problems.