Papers

12094 papers

ICML2025

Which Attention Heads Matter for In-Context Learning?

Summary pending...

interpretabilityin-context learninglarge language models
ICML2025

From Mechanistic Interpretability to Mechanistic Biology: Training, Evaluating, and Interpreting Sparse Autoencoders on Protein Language Models

Summary pending...

protein language modelsparse autoencodermechanistic interpretability
ICML2025

DMM: Distributed Matrix Mechanism for Differentially-Private Federated Learning Based on Constant-Overhead Linear Secret Resharing

Summary pending...

federated learningdifferential privacymatrix mechanism
ICML2025

BEST-Route: Adaptive LLM Routing with Test-Time Optimal Compute

Summary pending...

Large language modelsEfficient MLQuery Routing
ICML2025

Auditing $f$-differential privacy in one run

Summary pending...

Empirical privacyAuditingDifferential Privacy
ICML2025

Improved Algorithm for Deep Active Learning under Imbalance via Optimal Separation

Summary pending...

Deep LearningActive LabelingClass Imbalance
ICML2025

Probabilistic Factorial Experimental Design for Combinatorial Interventions

Summary pending...

Experimental DesignFactorial ExperimentCombinatorial Interventions
ICML2025

CoastalBench: A Decade-Long High-Resolution Dataset to Emulate Complex Coastal Processes

Summary pending...

Coastal ProcessesDatasetBenchmark
ICML2025

Causal-PIK: Causality-based Physical Reasoning with a Physics-Informed Kernel

Summary pending...

active explorationphysical reasoningcausality
ICML2025

Emergent Symbolic Mechanisms Support Abstract Reasoning in Large Language Models

Summary pending...

Mechanistic interpretabilityreasoningabstraction
ICML2025

Multiple-policy Evaluation via Density Estimation

Summary pending...

Theoretical Reinforcement Learning; Policy Evaluation; Instance dependent results;
ICML2025

EmbodiedBench: Comprehensive Benchmarking Multi-modal Large Language Models for Vision-Driven Embodied Agents

Summary pending...

Embodied AgentMulti-modal Large Language Models
ICML2025

Joint Localization and Activation Editing for Low-Resource Fine-Tuning

Summary pending...

activation editinglow-resourcelocalization
ICML2025

Stronger Neyman Regret Guarantees for Adaptive Experimental Design

Summary pending...

ATEtreatment effectadaptive experimental design
ICML2025

Symmetric Reinforcement Learning Loss for Robust Learning on Diverse Tasks and Model Scales

Summary pending...

Reinforcement learningCross entropySymmetric loss functions
ICML2025

Learning with Expected Signatures: Theory and Applications

Summary pending...

Probabilistic Machine LearningSignatureExpected Signature
ICML2025

Federated Generalised Variational Inference: A Robust Probabilistic Federated Learning Framework

Summary pending...

Federated LearningProbabilistic Machine LearningModel Misspecification
ICML2025

Boosting Protein Graph Representations through Static-Dynamic Fusion

Summary pending...

Graph Neural NetworksProtein ModelingMolecular Dynamics
ICML2025

Simplifying DINO via Coding Rate Regularization

Summary pending...

Self-supervised learningDINO
ICML2025

Wasserstein Policy Optimization

Summary pending...

Policy OptimizationWasserstein metricOptimal Transport