arXiv:2604. 16557v2 Announce Type: replace Abstract: Current post-training methodologies for adapting Large Vision-Language Models (LVLMs) generally fall into two paradigms: Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL).
The paper introduces Objective-aware Trajectory Credit Assignment (OTCA), a framework that refines reinforcement learning for diffusion-based visual generation. OTCA decomposes credit across denoising steps and allocates multiple reward signals adaptively, addressing the coarse, uniform reward assignment of existing GRPO pipelines. Experiments demonstrate that OTCA consistently enhances image and video generation quality across various metrics.
arXiv:2607. 04713v1 Announce Type: cross Abstract: Reinforcement learning holds significant potential for training large language models (LLMs) to handle multi-turn interactive tasks.
arXiv:2506. 13702v4 Announce Type: replace-cross Abstract: Single-trajectory preference optimization methods learn from datasets of ((prompt, response, reward)) tuples, offering a practical alternative to pairwise preference learning by directly leveraging scalar feedback.
The paper tackles sample efficiency in image-based reinforcement learning by combining novelty and surprise signals to prioritize experiences. It proposes Novelty and Surprise Prioritized Experience Replay (NSPER) and an extended version, NSPER+R, which also uses these signals as intrinsic rewards. Experiments on DeepMind Control Suite tasks demonstrate that both methods accelerate training and improve convergence compared to existing techniques.
Prism‑GRPO enhances the GRPO reinforcement‑learning algorithm for vision‑language‑action policies by adding a weighted trajectory‑level execution‑quality score to binary success rewards. This approach splits groups with identical outcomes into a quality spectrum, preserving training signal while ensuring successes always outrank failures. Experiments on four RoboTwin tasks show Prism‑GRPO achieves higher success and quality at matched rollout budgets, reaching target success rates with up to 56% fewer rollouts and mitigating reward‑hacking behaviors that transfer to real‑robot deployment.
Routing Policy Optimization (RPO) introduces stochastic exploration within a diffusion visual generator by treating attention routing as a stochastic action. Instead of perturbing the entire reverse denoising trajectory, RPO samples key attention blocks at selected denoising levels, perturbs the transition velocity, and evaluates the perturbation with a score‑function objective. A two‑rollout estimator provides an advantage signal, and after training the model restores dense attention, yielding higher generation quality and nearly four times the training efficiency compared to Flow‑GRPO‑style methods.
The paper proposes Novelty and Surprise Prioritized Experience Replay (NSPER) for image-based reinforcement learning, combining novelty to highlight underrepresented states and surprise to reveal gaps in the agent’s knowledge. An extended version, NSPER+R, also uses these signals as intrinsic rewards to enhance both replay quality and exploration. Experiments on DeepMind Control Suite tasks demonstrate that NSPER and NSPER+R accelerate training and improve convergence compared to existing methods.
arXiv:2608.23664v1 Announce Type: cross Abstract: Reward fine-tuning is becoming an important tool for adapting diffusion models to human preferences and task-specific objectives, but existing method...
The paper proposes a reward-based policy that relies only on rewards and actions, enabling zero‑shot transfer between source and target environments with entirely different observation spaces. Experiments on Pointmass, Cartpole, 2D Car Racing, and the Stretch robot in Habitat‑Sim show that the policy can adapt to new visual styles or 3D renderings without additional samples. Additionally, the reward policy can guide the training of an observation‑based policy in the target environment.
The paper introduces Reflection-Aware GRPO (RA‑GRPO), a reinforcement‑learning framework that aligns diffusion generative models with human preferences. It uses Diffusion Reflection to correct intermediate sampling paths by reversing the diffusion process, and Counterfactual Path Synthesis to embed these corrected trajectories into the policy, avoiding extra inference cost. Experiments on text‑to‑image and text‑to‑video models show RA‑GRPO outperforms existing methods, reducing reward hacking and improving generalization while remaining architecture‑agnostic.
arXiv:2603. 23086v2 Announce Type: replace Abstract: Autoregressive (AR) models are highly effective for image generation, yet their standard maximum-likelihood estimation training lacks direct optimization for sample quality and diversity.