Opening the Black Box of Classifier-Free Guidance via Information Bottleneck
cs.LG
Submitted: 2026-10-09
Updated: 2026-10-09
Sources
- Deep Unsupervised Learning using Nonequilibrium Thermodynamics
- Denoising Diffusion Implicit Models
- Score-Based Generative Modeling through Stochastic Differential Equations
- Diffusion Models Beat GANs on Image Synthesis
- Classifier-Free Diffusion Guidance
- eDiff-I: Text-to-Image Diffusion Models with an Ensemble of Expert Denoisers
- Hierarchical Text-Conditional Image Generation with CLIP Latents
- Classifier-free Guidance with Adaptive Scaling
- CFG-Ctrl: Control-Based Classifier-Free Diffusion Guidance
- Guidance in the Frequency Domain Enables High-Fidelity Sampling at Low CFG Scales
- Token Perturbation Guidance for Diffusion Models
- Scalable Diffusion Models with Transformers
- The information bottleneck method
- Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow
- Flow Matching for Generative Modeling
- Denoising Diffusion Probabilistic Models
- Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis
- Momentum Guidance: Plug-and-Play Guidance for Flow Models
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks