MIRA: Mutual Information guided calibration for Reliable Test-Time Adaptation
MIRA is a source-free online adaptation method using lightweight Gaussian perturbations and mutual information to improve test-time adaptation under distribution shift.
Stories per day · 30 days
4 RSS
Relatedcalibration1latent-divergence1learning-to-defer1mutual-information1offline-reinforcement-learning1optimization1policy-regularization1residual-analysis1robustness1test-time-adaptation1
15 stories, newest first · one card per story
Stories tagged #distribution-shift
Offline RL paper regularizes policies via consequence distributions
Structural Behavior Regularization replaces an intractable dynamics constraint with a closed-form latent divergence and reports stronger offline RL benchmark results.
REASON proposes two-loop residual analysis for learning
The two-loop model beats the fast loop at about half the medium loop's computation, including under unseen corruptions.
Why Linear Models Can Still Work in a Complex and Changing World
Linear models can beat nonlinear ones when unstable curvature fails to transfer across environments.
Density Ratio Estimation with Stein Displacement Fields
Stein displacement fields estimate density ratios and distribution shift through one convex problem.
PairAudit: Guiding Human Review with Graph Tokens under Distribution Shift
PairAudit uses graph tokens to surface overlooked intrusion-detector mistakes and improve fixed-budget review of unseen attacks.
False Floors: LLM Safety Routing Evaluations Break Under Distribution Shift
Safety-router benchmarks overstate routing gains by picking the best model on test data.
Attention Function as an Intrinsic Inductive Bias: How Models' Behavior Diverges in Novel Contexts
A fixed softmax-sigmoid attention mix barely changes GPT-2 in-distribution loss but shifts out-of-distribution perplexity sharply.
WISE-ATTA: When to Ask for Labels in Budgeted Active Test-Time Adaptation
WISE-ATTA schedules scarce labels during test-time adaptation and matches prior methods with fewer annotations.
A Unified Uncertainty Representation for Graph Neural Networks via Doubly-Spectral Stochastic Expansion
DSS-GNN unifies GNN calibration, OOD detection, and shift robustness in one spectral representation.
PyroAdapt: Adapting Wildfire Prediction under Spatial Heterogeneity and Temporal Shift
PyroAdapt adapts wildfire models by retrieving similar sites and ranking fire-prone cells under a daily budget.
Online Learning via Learned Latent Bayesian Tracking
AURA learns a latent state space so deep models adapt online via Kalman filtering under distribution shift.
Partition-Matched Evaluation of Community Features under Distribution Shift in Android Malware Function-Call Graphs
Study finds Android malware graph communities add little shift-stable signal beyond random partitions.
How Does Distribution Shift Shape Pretraining Gains in Neural PDE Surrogates?
Study shows pretraining gains for neural PDE surrogates depend on distribution shift type, target data budget, and whether source-target physics differ.
General Quantification of Covariate and Concept Shifts
Paper proposes γ*-concept shifts via entropic optimal transport, deriving estimable generalization bounds unifying covariate and concept shift under distribution shift.