MCLR: Improving Conditional Modeling via Inter-Class Likelihood-Ratio Maximization and Unifying Classifier-Free Guidance with Alignment Objectives

arXiv:2603.22364 · cs.LG, cs.AI, cs.CV · Submitted 2026-03-23 · Read on arXiv

cs.LG, cs.AI, cs.CV

Submitted: 2026-03-23

Updated: 2026-10-20

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Diffusion Model — 3× in this paper · explained in 10 episodes
• Diffusion Models — 3× in this paper · explained in 9 episodes
• Contrastive Alignment — 1× in this paper · explained in 2 episodes
• Fine-Tuning — 1× in this paper · explained in 5 episodes
• Generative Modeling — 1× in this paper · explained in 3 episodes
• Inference-Time Guidance — 1× in this paper · explained in 3 episodes
• Training Objective — 1× in this paper · explained in 3 episodes

Sources

• Classifier-Free Guidance is a Predictor-Corrector
• Diffusion Blend: Inference-Time Multi-Preference Alignment for Diffusion Models
• Diffusion Guidance Is a Controllable Policy Improvement Operator
• Classifier-Free Diffusion Guidance
• Inference-Time Alignment Control for Diffusion Models with Reinforcement Learning Guidance
• Stage-wise Dynamics of Classifier-Free Guidance in Diffusion Models
• Rethinking Preference Alignment for Diffusion Models with Classifier-Free Guidance
• Feature-guided score diffusion for sampling conditional densities
• Aligning Text to Image in Diffusion Models is Easier Than You Think
• Representation Learning with Contrastive Predictive Coding
• Classifier-Free Guidance: From High-Dimensional Analysis to Generalized Guidance Forms
• Hierarchical Text-Conditional Image Generation with CLIP Latents
• Diffusion Models without Classifier-free Guidance
• Emergence of Distortions in High-Dimensional Guided Diffusion Models
• No Alignment Needed for Generation: Learning Linearly Separable Representations in Diffusion Models
• Training Class-Imbalanced Diffusion Model Via Overlap Optimization
• Direct Discriminative Optimization: Your Likelihood-Based Visual Generative Model is Secretly a GAN Discriminator

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