Generator-Aligned Reward Modeling with Target-Aware Bounded Optimization for Visual Generation
cs.CV
Submitted: 2026-11-05
Updated: 2026-11-06
Terminology
Sources
- VAST 1.0: A Unified Framework for Controllable and Consistent Video Generation
- Training Diffusion Models with Reinforcement Learning
- Video Diffusion Alignment via Reward Gradients
- Flow-GRPO: Training Flow Matching Models via Online RL
- Learning to Credit the Right Steps: Objective-aware Process Optimization for Visual Generation
- Reward-Aware Trajectory Shaping for Few-step Visual Generation
- Scaling Laws for Reward Model Overoptimization
- ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generation
- Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generation
- Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis
- VideoScore: Building Automatic Metrics to Simulate Fine-grained Human Feedback for Video Generation
- VisionReward: Fine-Grained Multi-Dimensional Human Preference Learning for Image and Video Generation
- Improving Video Generation with Human Feedback
- VideoScore2: Think before You Score in Generative Video Evaluation
- Video Generation Models Are Good Latent Reward Models
- Aligning Text-to-Image Diffusion Models with Reward Backpropagation
- DanceGRPO: Unleashing GRPO on Visual Generation
- A Systematic Post-Train Framework for Video Generation
- TeleBoost: A Systematic Alignment Framework for High-Fidelity, Controllable, and Robust Video Generation
- GRPO-Guard: Mitigating Implicit Over-Optimization in Flow Matching via Regulated Clipping
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs