Tag
The paper proposes Scout, a framework for budgeted task-aware acquisition of dynamic networks that learns to prioritize graph updates based on downstream task value, achieving superior performance over existing methods in multiple benchmarks.
This paper introduces HybridAL, a method for active learning that dynamically switches between retraining and fine-tuning based on a stabilization signal, improving efficiency and performance in text classification tasks.
DE-Venus is a unified framework for data-efficient Reinforcement Learning with Verifiable Rewards in large language models, reducing annotation and training costs while maintaining performance.
A physics-informed graph attention network is proposed as a surrogate for TCAD simulation, enabling faster design space exploration by predicting mesh-node values and incorporating carrier-transport physics.
AdaptNTK introduces an adaptive uncertainty quantification framework using neural tangent kernels to enable efficient active learning in neural network potentials, reducing computational cost while maintaining accuracy in molecular dynamics simulations.
This paper introduces PATH, a framework for active curriculum refinement in reinforcement learning that leverages curriculum graph structure to improve robustness and generalization across diverse environments.
This controlled study investigates whether open-weight LLMs can serve as acquisition policies for finite-pool materials optimization, comparing their performance with random selection and conventional Gaussian-process methods across multiple tasks.
This paper studies federated active learning in low-budget regimes, revealing that homogeneous data requires stronger coordination due to heterogeneity reversal. It proposes a framework using federated representation learning to enable globally coordinated active selection, outperforming existing methods.
The paper introduces XIGL, a human-in-the-loop strategy using explanations and active learning to detect and remove shortcuts in Graph Neural Networks, enhancing their reliability for out-of-distribution tasks.
This study tests uncertainty sampling in active learning under bounded label noise, comparing error exposure and location effects across datasets to assess robustness and performance.
The paper tests the adage that financial news is priced in before publication, using LLM distillation to classify news events and analyze market reactions. It finds that price moves concentrate before and at publication, with different patterns for fundamental versus story-driven news.
Introduces Adversarial Causal Intervention Falsification (ACIF), a sequential game where a structural causal generator proposes observational and interventional distributions while an adversarial experimentalist selects interventions to falsify it. The paper provides theoretical guarantees, including finite-sample convergence and model-selection, bridging causal generative modeling, active discovery, and experimental design.
This paper proposes a unified optimal transport framework for cold-start active learning, introducing a Sinkhorn-based algorithm (ε-AS) that adapts regularization strength to data and achieves state-of-the-art results on six datasets, including improving ImageNet-1k accuracy by 1.29% over prior methods while reducing selection time by 56.2%.
This paper addresses the fixed-confidence top-k identification problem from noisy pairwise comparisons, and develops an asymptotically optimal algorithm that minimizes the expected number of comparisons.
Introduces ZendoWorld, a controlled interactive environment for evaluating AI agents on active visual concept induction, where agents must perceive scenes, infer hidden logical rules, and design informative experiments. Experiments with various agent classes reveal that high prediction accuracy does not guarantee rule recovery, and VLM-based agents struggle with informative experimentation, highlighting gaps compared to human inductive reasoning.
Proves a tight approximation ratio for the greedy algorithm in myopic Bayesian active learning for linear regression, identifying the maximum initial leverage score as a key quantity.
Proposes a novel framework combining active learning with masked reconstruction and minimax strategies to improve unsupervised time series anomaly detection, achieving 12.39% AUC improvement over baselines across 28 test cases.
This paper introduces Personalized Active Preference Alignment (PAPA), a method for fine-tuning diffusion models using real-time user feedback without a parameterized reward model, enhancing efficiency in personalized tasks like recommendations and image generation.
This paper proposes active shot allocation strategies for quantum kernel estimation in Gaussian process regression, deriving pair-level sensitivities to guide non-uniform shot budgets and demonstrating significant improvements in test RMSE over uniform allocation on benchmarks.
This paper introduces reinforcement learning with metacognitive feedback (RLMF) and metacognitive data selection to improve large language model calibration, enabling faithful expression of intrinsic uncertainty and surpassing standard RL by up to 63%.