Research map / Reinforcement learning / Policy gradient and actor-critic
Reward-Directed Score-Based Diffusion Models via q-Learning
Xuefeng Gao, Jiale Zha, Xun Yu Zhou
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| increase score-estimator sample size m from 30 to 120 (ablation) | q-learning diffusion · Swiss roll (beta=0) | KL divergence to data: 0.33 -> 0.17 | Improved |
| increase sample size m from 30 to 150 at fixed reward 0.8 (ablation) | q-learning diffusion · Swiss roll (reward 0.8) | KL divergence to data: 0.88 -> 0.38 | Improved |
| increase sample size m from 150 to 300 (ablation) | q-learning diffusion · Swiss roll (beta=0) | KL divergence to data: 0.18 -> 0.17 | No clear effect |
| increase sample size m from 30 to 300 (t=1) (ablation) | ratio score estimator · Swiss roll, t=1 | MSE of score estimate: 5.348×10^-2 -> 8.682×10^-3 | Improved |
| increase sample size m from 30 to 300 (t=4) (ablation) | ratio score estimator · Swiss roll, t=4 | MSE of score estimate: 1.112×10^-3 -> 3.209×10^-4 | Improved |
| increase sample size m from 30 to 300 (t=1) (ablation) | ratio score estimator · Swiss roll, t=1 | variance of score estimate: 5.036×10^-2 -> 4.458×10^-3 | Improved |
| evaluate at larger t (t=4 instead of t=1) | ratio score estimator · Swiss roll, m=30 | MSE of score estimate: 5.348×10^-2 -> 1.112×10^-3 | Improved |
Closest papers by meaning
- Score as Action: Fine Tuning Diffusion Generative Models by Continuous-time Reinforcement Learning (ICML 2025 · similarity 0.92)
- Learning a Diffusion Model Policy from Rewards via Q-Score Matching (ICML 2024 · similarity 0.91)
- Continuous Q-Score Matching: Diffusion Guided Reinforcement Learning for Continuous-Time Control (NeurIPS 2025 · similarity 0.90)
- Diffusion-based Reinforcement Learning via Q-weighted Variational Policy Optimization (NeurIPS 2024 · similarity 0.90)
- Designing Reinforcement Learning for Diffusion Models: A Unified Path-Space View (NeurIPS 2026 · similarity 0.88)
- Diffusion Policies Creating a Trust Region for Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
