Do Visual RL Gains Scale at Test Time? A Quantile View of Post-Training
cs.LG, cs.CV
Submitted: 2026-11-02
Updated: 2026-11-02
Code: https://github.com/austin-k-wang/QUPO
Terminology
Sources
- Qwen3-VL Technical Report
- How RL Reshapes LLM Reasoning: Transferability, Coverage, and Scaling Laws
- Rethinking the Design Space of Reinforcement Learning for Diffusion Models: On the Importance of Likelihood Estimation Beyond Loss Design
- Emerging Properties in Unified Multimodal Pretraining
- How to Guide Your Flow: Few-Step Alignment via Flow Map Reward Guidance
- GenEval 2: Addressing Benchmark Drift in Text-to-Image Evaluation
- What should post-training optimize? A test-time scaling law perspective
- Flow Matching for Generative Modeling
- UniGRPO: Unified Policy Optimization for Reasoning-Driven Visual Generation
- Smoothed Preference Optimization via ReNoise Inversion for Aligning Diffusion Models with Varied Human Preferences
- Multi-Axis Max@K Reinforcement Learning for Representative Diversity in Text-to-Image Generation
- Tail-Likelihood Reinforcement Learning
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- A General Framework for Inference-time Scaling and Steering of Diffusion Models
- Maximum Likelihood Reinforcement Learning
- On Advantage Estimates for Max@K Policy Gradients
- Inference-Time Alignment in Diffusion Models with Reward-Guided Generation: Tutorial and Review
- Aligning One-Step Generative Models with Reward-Weighted Transport Distillation
- Beyond Pairwise Preferences: Listwise Reward-Aware Alignment for Diffusion Models
- InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis