Research map / Image generation / Fast diffusion sampling

Generative Active Learning via Bayesian Acquisition for Improving the Efficiency of Synthetic Data

NeurIPS 2026 · Image generation › Fast diffusion sampling · approach: Distributional control of diffusion

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes steering diffusion models with Bayesian epistemic uncertainty (BALD-based acquisition) so that generated synthetic training data is more informative for the target model.

Problem

Synthetic data from generative models is not automatically informative for training target models, so its training value must be explicitly considered.

Approach

Uses Bayesian epistemic uncertainty to guide diffusion model sampling toward informative samples, estimating BALD scores in a computationally tractable way to act as an acquisition function for generative active learning.

Experimental setup

Builds on: BALD

Closest papers by meaning

• Deep Bayesian Active Learning for Preference Modeling in Large Language Models (NeurIPS 2024 · similarity 0.86)
• Transductive Active Learning: Theory and Applications (NeurIPS 2024 · similarity 0.86)
• The Unmet Promise of Synthetic Training Images: Using Retrieved Real Images Performs Better (NeurIPS 2024 · similarity 0.84)
• Epistemic Uncertainty for Generated Image Detection (NeurIPS 2025 · similarity 0.84)
• Hyperbolic Active Learning for Semantic Segmentation under Domain Shift (ICML 2024 · similarity 0.84)
• BridgeTwist: Twisting Schrödinger Bridges for Training-Free Conditional Sampling (NeurIPS 2026 · similarity 0.84)