Self-Verifier Capture in Unified Visual Generators: Diagnosing and Repairing Closed-Loop Self-Reward
cs.CV
Submitted: 2026-10-12
Updated: 2026-10-12
Comments: 36 pages, 5 figures; supplementary material included
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- IRIS: Intrinsic Reward Image Synthesis
- ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignment
- LatentUM: Unleashing the Potential of Interleaved Cross-Modal Reasoning via a Latent-Space Unified Model
- VA-$\pi$: Variational Policy Alignment for Pixel-Aware Autoregressive Generation
- Flow-GRPO: Training Flow Matching Models via Online RL
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- Breaking the Self-Confirming Loop: Diagnosing and Mitigating Systemic Reward Bias in Self-Rewarding RL
- Scaling Text-to-Image Diffusion Transformers with Representation Autoencoders
- SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features
- Reward Model Overoptimisation in Iterated RLHF
- Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis
- Qwen2.5 Technical Report
- KVPO: ODE-Native GRPO for Autoregressive Video Alignment via KV Semantic Exploration
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models