Research map / Reinforcement learning / Visual and representation RL
Discrete Compositional Generation via General Soft Operators and Robust Reinforcement Learning
Marco Jiralerspong, Esther Derman, Danilo Vucetic, Nikolay Malkin, Bilun Sun, Tianyu Zhang, Pierre-Luc Bacon, Gauthier Gidel
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 |
|---|---|---|---|
| use trajectory general mellowmax (q=0.75) instead of GFN | GFN · GFP | average mode reward: 1.90 -> 2.95 | Improved |
| use TGM q=0.75 instead of GFN | GFN · AMP | average mode reward: 10.04 -> 10.43 | Improved |
| use TGM q=0.75 instead of GFN | GFN · UTR | average mode reward: 4.12 -> 4.27 | Improved |
| raise q from 0.25 to 1.0 in TGM | TGM · GFP | average mode reward: 2.95 -> 2.05 | Hurt |
| use q=1.0 instead of q=0.75 in TGM | TGM · AMP | average mode reward: 10.43 -> 10.29 | No clear effect |
| use TGM instead of PPO | PPO · GFP | average mode reward: 0.66 -> 2.95 | Improved |
| use TGM instead of SAC | SAC · AMP | average mode reward: 5.69 -> 10.43 | Improved |
| use TGM/TB instead of PPO | PPO · GFP | training speed (samples/s): 8 -> 120 | Improved |
| use TGM instead of PPO | PPO · AMP | average mode reward: 9.79 -> 10.13 | No clear effect |
Closest papers by meaning
- Aligning Transformers with Continuous Feedback via Energy Rank Alignment (NeurIPS 2025 · similarity 0.87)
- Improved Off-policy Reinforcement Learning in Biological Sequence Design (ICML 2025 · similarity 0.87)
- Feedback Efficient Online Fine-Tuning of Diffusion Models (ICML 2024 · similarity 0.87)
- Flow Density Control: Generative Optimization Beyond Entropy-Regularized Fine-Tuning (NeurIPS 2025 · similarity 0.87)
- Geometric Active Exploration in Markov Decision Processes: the Benefit of Abstraction (ICML 2024 · similarity 0.86)
- Derivative-Free Guidance in Continuous and Discrete Diffusion Models with Soft Value-based Decoding (NeurIPS 2025 · similarity 0.86)
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.
