archive
Every paper Pith has read. Search by title, abstract, or pith.
14903 papers in cs.LG · page 6
-
RL cuts ion shuttling operations by 36 percent
Reinforcement learning for ion shuttling on trapped-ion quantum computers
-
Synthetic RAW data yields same low-light detection metrics as real
Making the Discrete Continuous: Synthetic RAW Augmentations for Fine-Grained Evaluation of Person Detection Performance in Low Light
-
Rehearsal of stored Q-values cuts forgetting in repeated RL tasks
Don't Forget the Critic: Value-Based Data Rehearsal for Multi-Cyclic Continual Reinforcement Learning
-
Algorithms achieve optimal bidding rates despite feedback shilling
Do Not Trust The Auctioneer: Learning to Bid in Feedback-Manipulated Auctions
-
EM pulses trigger persistent accuracy collapse on NCS2 until reload
Characterizing the Fault Response of the Intel Neural Compute Stick 2 Under Single-Pulse Electromagnetic Fault Injection
-
AMUSE blends Muon speed with schedule-free stability without learning rate schedules
AMUSE: Anytime Muon with Stable Gradient Evaluation
-
Separate physical pools for KV and SSM caches cut OOMs 7.6% and raise throughput up to 13x
Asymmetric Virtual Memory Paging for Hybrid Mamba-Transformer Inference
-
Query-time RL turns noisy memory into accurate evidence
DeferMem: Query-Time Evidence Distillation via Reinforcement Learning for Long-Term Memory QA
-
MDL granular-ball tree regularizes spectral clustering graphs
Minimum Description Length based Granular-Ball Tree Regularization for Spectral Clustering
-
Early visual alignment conserved across species but IT rankings diverge
Cross-Species RSA Reveals Conserved Early Visual Alignment but Divergent Higher-Area Rankings Across Human fMRI and Macaque Electrophysiology
-
Law of total variance splits wind forecast uncertainty
A Posterior-Predictive Variance Decomposition for Epistemic and Aleatoric Uncertainty in Wind Power Forecasting
-
Hybrid model cuts NEM electricity price forecast errors by 12%
Hybrid Kolmogorov-Arnold Network and XGBoost Framework for Week-Ahead Price Forecasting in Australia's National Electricity Market
-
Message passing makes GNN-LRP subgraph scores linear in depth
Efficient Higher-order Subgraph Attribution via Message Passing
-
Multi-stage pipeline cuts false positives in Indic abusive comment detection
Multi-Stage Training for Abusive Comment Detection in Indic Languages
-
Local EEG segment matching raises cross-subject emotion accuracy
Cross-Subject EEG Emotion Recognition Based on Temporal Asynchronous Alignment Contrastive Learning
-
BERT layer 8 activates on content words over structure
Towards Explainability of SLMs by investigating Token Level Activation
-
Bellman alignment selects better source data than transition similarity
Target-Aligned Bellman Backup for Cross-domain Offline Reinforcement Learning
-
Boundary attacks recover 19% of safety classifier training data
Boundary-targeted Membership Inference Attacks on Safety Classifiers
-
Attack recovers 19% of safety classifier distress data
Boundary-targeted Membership Inference Attacks on Safety Classifiers
-
Attention sink pruning accelerates Vision Transformers by 48%
ASAP: Attention Sink Anchored Pruning
-
Adversarial scaling reveals LLM code weaknesses
VeriScale: Adversarial Test-Suite Scaling for Verifiable Code Generation
-
TimeGuard boosts backdoor resistance in time series forecasts by 1.96x
TimeGuard: Channel-wise Pool Training for Backdoor Defense in Time Series Forecasting
-
LLMs learn to plan transit routes from records alone
TransitLM: A Large-Scale Dataset and Benchmark for Map-Free Transit Route Generation
-
Selective neuron fusion trades ensemble accuracy for lower cost
Partial Fusion of Neural Networks: Efficient Tradeoffs Between Ensembles and Weight Aggregation
-
Regular graphs make ASE and LSE subspaces identical
The ASE-LSE Disagreement Landscape: An End-to-End Characterisation of Extremes and Structural Drivers
-
Boundary layers drive one-third power-law error decay in online softmax training
A Boundary-Layer Mechanism for One-Third Scaling in Online Softmax Classification
-
Flow matching reconstructs cell trajectories from snapshots
From Snapshots to Trajectories: Learning Single-Cell Gene Expression Dynamics via Conditional Flow Matching
-
Generative solver enforces conservation laws during sampling
Physics-Informed Generative Solver: Bridging Data-Driven Priors and Conservation Laws for Stable Spatiotemporal Field Reconstruction
-
Learned causal order improves tabular predictions after interventions
Learning Causal Orderings for In-Context Tabular Prediction
-
Off-log map turns fMRI correlations into standard statistics
Riemannian geometry meets fMRI: the advantages of modeling correlation manifolds and eigenvector subspaces
-
Scaling sepsis AI replicas to CPU threads halves detection latency
SepsisAI Orchestrator: A Containerized and Scalable Platform for Deploying AI Models and Real-Time Monitoring in Early Sepsis Detection
-
Chebyshev policies solve Mountain Car optimally with 277x fewer parameters
Chebyshev Policies and the Mountain Car Problem: Reinforcement Learning for Low-Dimensional Control Tasks
-
Benchmark compares 12 pipelines for knowledge graph integration
Evaluation of Pipelines for Data Integration into Knowledge Graphs
-
Benchmarks reveal when AI coordination improves scientific inference
Cross-domain benchmarks reveal when coordinated AI agents improve scientific inference from partial evidence
-
Heavy-tailed spectra set per-layer rates to speed LLM training 1.5x
One LR Doesn't Fit All: Heavy-Tail Guided Layerwise Learning Rates for LLMs
-
Decomposition restores long-term fairness despite selective labels
Long-term Fairness with Selective Labels
-
EmoTrack cuts depression prediction error 13.5% on single sessions
EmoTrack: Robust Depression Tracking from Counseling Transcripts across Session Regimes
-
Adaptive allocation of noisy kernel entries improves SVM accuracy
Adaptive Measurement Allocation for Learning Kernelized SVMs Under Noisy Observations
-
Probe set partitions flag atypical federated clients
Detecting Atypical Clients in Federated Learning via Representation-Level Divergence
-
Adaptive distillation preserves exploration for better LLM math scores
Tailoring Teaching to Aptitude: Direction-Adaptive Self-Distillation for LLM Reasoning
-
Audio denoiser infers scene to keep relevant sounds
Automatic Contextual Audio Denoising
-
Evidence hierarchy lifts Bayesian threat classification to 95%
An Evidence Hierarchy for Bayesian Object Classification via OSINT-Aided Heterogeneous Sensor Fusion
-
Rewriting equivalents at test time stabilizes LLM theorem proving
What are the Right Symmetries for Formal Theorem Proving?
-
Physical decompositions boost climate emulator OOD performance
No Epoch Like the Present: Robust Climate Emulation Requires Out-of-Distribution Generalisation
-
AI recommender lifts Cox fall-risk model C-index from 0.805 to 0.815
Explainable AI for Data-Driven Design of High-Dimensional Predictive Studies
-
Perturbations regulate decorrelation rates in Lorenz '96 ensembles
Decomposing Ensemble Spread in Lorenz '96 With Learned Stochastic Parameterizations
-
Quadratic ReLU replacement keeps decisions intact for FHE inference
Decision-Aware Quadratic ReLU Replacement for HE-Friendly Inference
-
Quadratic ReLU replacement preserves calibration decisions
Decision-Aware Quadratic ReLU Replacement for HE-Friendly Inference
-
KAN neural ODE recovers exact symbolic equations for holomorphic fractals
Holomorphic Neural ODEs with Kolmogorov-Arnold Networks for Interpretable Discovery of Complex Dynamics
-
Transformer output sequences grow linearly with prompt length
How Many Different Outputs Can a Transformer Generate?