Jeongjae Lee

Jaewoo Lee, Hyeongyu Kang, Dohyun Kim et al.

Aligning a few-step generative model is challenging, since existing alignment frameworks typically rely on restrictive assumptions: a tractable likelihood, a specific ODE/SDE solver, or a particular model family. We introduce FAV, Few-step Generative Models Alignment via Sample-based Variational Inference, a general alignment framework that requires only sample access to the generator and the reference distribution. We cast alignment as sampling from a reward-tilted distribution anchored to a reference distribution. We leverage Stein Variational Gradient Descent as a sample-based variational inference scheme and amortize its particle updates into the generator parameters via fixed-point regression. We evaluate FAV on two domains: robotics manipulation and image generator alignment. On generative policy alignment for robotic manipulation, FAV outperforms prevailing policy extraction baselines across 56 offline and 30 offline-to-online RL tasks. For image generator alignment, FAV fine-tunes diverse few-step backbones, including GAN, drifting model, consistency models, and flow maps, scaling from ImageNet-$256$ to 1024$^2$ text-to-image synthesis. Code is available at https://github.com/Jaewoopudding/FAV.

Jeongjae Lee, Jinho Chang, Jeongsol Kim et al.

Reward-based fine-tuning aims to steer a pretrained diffusion or flow-based generative model toward higher-reward samples while remaining close to the pretrained model. Although existing methods are motivated by different perspectives such as Soft RL, GFlowNets, etc., we show that many can be written under a common framework, which we call reward score matching (RSM). Under this view, alignment becomes score matching toward a reward-guided target, and the main differences across methods reduce to the construction of the value-guidance estimator and the effective optimization strength across timesteps. This unification clarifies the bias--variance--compute tradeoffs of existing designs and distinguishes core optimization components from auxiliary mechanisms that add complexity without clear benefit. Guided by this perspective, we develop simpler redesigns that improve alignment effectiveness and compute efficiency across representative settings with differentiable and black-box rewards. Overall, RSM turns a seemingly fragmented collection of reward-based fine-tuning methods into a smaller, more interpretable, and more actionable design space.

Jeongjae Lee, Jong Chul Ye

While reinforcement learning has advanced the alignment of text-to-image (T2I) models, state-of-the-art policy gradient methods are still hampered by training instability and high variance, hindering convergence speed and compromising image quality. Our analysis identifies a key cause of this instability: disproportionate credit assignment, in which the mathematical structure of the generative sampler produces volatile and non-proportional feedback across timesteps. To address this, we introduce Proportionate Credit Policy Optimization (PCPO), a framework that enforces proportional credit assignment through a stable objective reformulation and a principled reweighting of timesteps. This correction stabilizes the training process, leading to significantly accelerated convergence and superior image quality. The improvement in quality is a direct result of mitigating model collapse, a common failure mode in recursive training. PCPO substantially outperforms existing policy gradient baselines on all fronts, including the state-of-the-art DanceGRPO.

Youngseo Cho, In Hee Kwak, Dohyeon Kim et al.

Bradykinesia, characterized by involuntary slowing or decrement of movement, is a fundamental symptom of Parkinson's Disease (PD) and is vital for its clinical diagnosis. Despite various methodologies explored to quantify bradykinesia, computer vision-based approaches have shown promising results. However, these methods often fall short in adequately addressing key bradykinesia characteristics in repetitive limb movements: "occasional arrest" and "decrement in amplitude." This research advances vision-based quantification of bradykinesia by introducing nuanced numerical analysis to capture decrement in amplitudes and employing a simple deep learning technique, LSTM-FCN, for precise classification of occasional arrests. Our approach structures the classification process hierarchically, tailoring it to the unique dynamics of bradykinesia in PD. Statistical analysis of the extracted features, including those representing arrest and fatigue, has demonstrated their statistical significance in most cases. This finding underscores the importance of considering "occasional arrest" and "decrement in amplitude" in bradykinesia quantification of limb movement. Our enhanced diagnostic tool has been rigorously tested on an extensive dataset comprising 1396 motion videos from 310 PD patients, achieving an accuracy of 80.3%. The results confirm the robustness and reliability of our method.