The minimax rate of estimating second-order calibration error is Õ(1/√n) with a matching Ω(1/√n) lower bound, enabled by analyticity from the sech kernel and yielding the first finite-sample guarantee for second-order Platt scaling.
hub Mixed citations
Bayesian Active Learning for Classification and Preference Learning
Mixed citation behavior. Most common role is background (67%).
40
Pith papers citing it
Background
67% of classified citations
abstract
Information theoretic active learning has been widely studied for probabilistic models. For simple regression an optimal myopic policy is easily tractable. However, for other tasks and with more complex models, such as classification with nonparametric models, the optimal solution is harder to compute. Current approaches make approximations to achieve tractability. We propose an approach that expresses information gain in terms of predictive entropies, and apply this method to the Gaussian Process Classifier (GPC). Our approach makes minimal approximations to the full information theoretic objective. Our experimental performance compares favourably to many popular active learning algorithms, and has equal or lower computational complexity. We compare well to decision theoretic approaches also, which are privy to more information and require much more computational time. Secondly, by developing further a reformulation of binary preference learning to a classification problem, we extend our algorithm to Gaussian Process preference learning.
hub tools
citation-role summary
background 5
method 1
citation-polarity summary
representative citing papers
Proves CLT for stochastic gradient non-reversible Langevin Monte Carlo and sufficient condition for variance reduction via anti-symmetric perturbation relative to reversible baseline.
ICL-derived intrinsic rewards are biased in general MDPs but asymptotically match true learning progress in non-temporal settings, with supporting experiments.
SymQNet uses offline RL to amortize acquisition for adaptive Hamiltonian learning, delivering 47-72x lower decision latency than online Bayesian baselines on Ising models while keeping posterior updates.
ATLAS uses active learning with disentangled RNN ensembles to design experiments that recover RL agent models from bandit behavior 5-10x more efficiently than random or expert baselines in simulations.
OVR is a new one-step lookahead BO algorithm with Monte Carlo approximations that achieves a vanishing Bayesian expected simple regret upper bound.
Introduces an active learning method using LOT and Gaussian Processes to select optimal timepoints for inferring continuous probability paths from sparse distributional snapshots.
Introduces LoRA-Curve parameterization to link independent LoRA optima via low-loss valleys, yielding higher predictive mutual information on reasoning and classification tasks with Qwen2.5 7B.
STAP reduces training data costs for PDE surrogates by selectively acquiring key time steps per trajectory instead of full simulations.
AB-SID-iVAR enables Gaussian process active learning for self-induced Boltzmann distributions by closed-form approximation of the target, with high-probability error vanishing guarantees and empirical gains on PES and drug discovery tasks.
LG-CoTrain, an LLM-guided co-training method, outperforms classical semi-supervised baselines for crisis tweet classification in low-resource settings with 5-25 labeled examples per class.
InfoChess proposes a symmetric adversarial game focused purely on information control and probabilistic king-location inference, with RL agents outperforming heuristic baselines and gameplay dissected via belief entropy, cross-entropy, and predictive scores.
A bias-aware Bayesian model with judge-specific covariates and a top-k membership uncertainty acquisition rule recovers accurate top-k rankings from noisy LLM judges using fewer comparisons than naive aggregation or standard active learning.
Introduces BAG, an open-source benchmark suite with baselines, datasets, and tasks for assessing Bayesian low-rank adaptation of multi-modal language models on calibration, robustness, and decision-making under uncertainty.
QUEST measures uncertainty via the Lebesgue volume of highest-density regions of a distribution's support, evaluated at robustness parameter alpha, and claims to satisfy UQ axioms while outperforming variance and differential entropy on selective prediction tasks.
Surprise-Guided MergeSort uses a VLM-based composite surprise scorer to prioritize human comparisons inside a MergeSort scheduler, skipping up to 535 pairs per session and raising Kendall's τ by 6-12 points over Active Elo at fixed budget across six benchmarks.
The mutual-information measure of epistemic uncertainty is not reducible by additional data, requiring a split into aleatoric, sample-reducible epistemic, and mechanism-reducible epistemic uncertainty.
ShaplEIG adaptively selects coalitions for Shapley value estimation via expected information gain from a Gaussian process surrogate, with closed-form computation and polynomial complexity, showing improved sample efficiency in low-budget regimes.
Active learning strategies for preference-based MPC objective learning achieve better closed-loop alignment with human preferences using fewer queries than random sampling in numerical tests.
Adaptive Prompt Elicitation (APE) uses an information-theoretic framework to generate visual queries that elicit and compile user intent into better prompts for text-to-image models, showing improved alignment in benchmarks and a user study.
UG-TTT adds epistemic uncertainty measured by adapter disagreement as an exploration bonus in RL for LLMs, raising maximum reward and diversity on scientific discovery benchmarks.
Decoupled PFNs use controllable synthetic priors to train separate latent-signal and noise heads, making epistemic-aleatoric decomposition identifiable and improving acquisition in noisy settings.
An ensemble-based information-theoretic active learning method using ensemble Kalman inversion selects valuable tasks to optimize communication structures in LLM multi-agent systems more reliably than random sampling under limited training budgets.
VPL learns individualized vibrotactile preferences efficiently via uncertainty-aware Gaussian process models and active query selection in a 13-participant user study on an Xbox controller.
citing papers explorer
No citing papers match the current filters.