Papers

12094 papers

ICML2025

MPO: An Efficient Post-Processing Framework for Mixing Diverse Preference Alignment

Summary pending...

Direct Preference OptimizationLarge Language ModelsRLHF
ICML2025

The underlying structures of self-attention: symmetry, directionality, and emergent dynamics in Transformer training

Summary pending...

Transformer modelsSelf-attentionDeep Learning Theory
ICML2025

Online Episodic Convex Reinforcement Learning

Summary pending...

Online learningconvex reinforcement learningmarkov decision processes
ICML2025

MultiPDENet: PDE-embedded Learning with Multi-time-stepping for Accelerated Flow Simulation

Summary pending...

physics-informed learningmultiscale time steppingspatiotemporal dynamics
ICML2025

Shortcut-connected Expert Parallelism for Accelerating Mixture of Experts

Summary pending...

Mixture of ExpertsExpert ParallelismShortcut Connection
ICML2025

Inverse Problem Sampling in Latent Space Using Sequential Monte Carlo

Summary pending...

Inverse problemssequential Monte CarloLatent diffusion
ICML2025

Spurious Correlations in High Dimensional Regression: The Roles of Regularization, Simplicity Bias and Over-Parameterization

Summary pending...

high-dimensional statisticsempirical risk minimizationspurious correlations
ICML2025

False Coverage Proportion Control for Conformal Prediction

Summary pending...

Conformal predictionmultiple testingFalse Discovery control
ICML2025

Multinoulli Extension: A Lossless Yet Effective Probabilistic Framework for Subset Selection over Partition Constraints

Summary pending...

Subset SelectionWeakly Submodular MaximizationPartition Matroid
ICML2025

Learning to Match Unpaired Data with Minimum Entropy Coupling

Summary pending...

Minimum entropy couplingUnsupervised learningDiffusion models
ICML2025

Prediction models that learn to avoid missing values

Summary pending...

Missing valuesinterpretablityboosting
ICML2025

Calibrated Value-Aware Model Learning with Probabilistic Environment Models

Summary pending...

model-based rlmuzeroitervaml
ICML2025

GANQ: GPU-Adaptive Non-Uniform Quantization for Large Language Models

Summary pending...

Large Language ModelsNon-uniform QuantizationGPU-adaptive Algorithms
ICML2025

Enhancing Diversity In Parallel Agents: A Maximum State Entropy Exploration Story

Summary pending...

Maximum State EntropyParallel Reinforcement LearningAgents' Diversity
ICML2025

Mind the Gap: a Spectral Analysis of Rank Collapse and Signal Propagation in Attention Layers

Summary pending...

transformersattention mechanismspectral analysis
ICML2025

HEAP: Hyper Extended A-PDHG Operator for Constrained High-dim PDEs

Summary pending...

PDENeural OperatorConstraints
ICML2025

InfoSEM: A Deep Generative Model with Informative Priors for Gene Regulatory Network Inference

Summary pending...

Gene Regulatory Networks inferenceInformative priorsscRNA-seq
ICML2025

Robust Autonomy Emerges from Self-Play

Summary pending...

Reinforcement LearningAutonomySimulation
ICML2025

Continuous-Time Analysis of Heavy Ball Momentum in Min-Max Games

Summary pending...

Min-Max GamesHeavy Ball MomentumContinuous-Time Analysis
ICML2025

A Variational Information Theoretic Approach to Out-of-Distribution Detection

Summary pending...

out-of-distribution detectioninformation theoryvariational calculus