Research map / Probabilistic and scientific modeling / Molecules and proteins
Spectral Flow Matching: Stabilizing Stochastic GFlowNets via Frequency-Domain Regularization
Nadhir Hassen, Johan W. Verjans
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 spectral loss (ST-GFN) instead of trajectory balance | TB · Sample efficiency | steps to 90% performance: 10483 -> 5850 | Improved |
| use spectral loss (ST-GFN) instead of SubTB | SubTB · Sample efficiency | steps to 90% performance: 7850 -> 5850 | Improved |
| use spectral loss (ST-GFN) instead of Stochastic-GFN | Stochastic-GFN · Sample efficiency | steps to 90% performance: 7680 -> 5850 | Improved |
| use spectral loss (ST-GFN) instead of TB+Novelty | TB+Novelty · Sample efficiency | steps to 90% performance: 8350 -> 5850 | Improved |
| use ST-GFN instead of TB | TB · Total wall-clock time | wall-clock time (s): 544s -> 404s | Improved |
| use ST-GFN instead of TB+Novelty | TB+Novelty · Total wall-clock time | wall-clock time (s): 548s -> 404s | Improved |
| add spectral regularization and RFF policy embedding | TB · BitSequence (90% action failure) | top-100 reward | Improved |
| add autocorrelated intrinsic reward | TB · HyperGrid 32x32 (64 modes) | modes found | Improved |
| add adaptive spectral regularization (learnable lambda) | ST-GFN · Training dynamics | optimization rate | Mixed |
Closest papers by meaning
- Stable GFlowNets with Probabilistic Guarantees (NeurIPS 2026 · similarity 0.91)
- Pre-Training and Fine-Tuning Generative Flow Networks (ICLR 2024 · similarity 0.89)
- Revisiting Non-Acyclic GFlowNets in Discrete Environments (ICML 2025 · similarity 0.89)
- Flow Factorization for Efficient Generative Flow Networks (AAAI 2025 · similarity 0.89)
- Expected flow networks in stochastic environments and two-player zero-sum games (ICLR 2024 · similarity 0.89)
- Beyond Squared Error: Exploring Loss Design for Enhanced Training of Generative Flow Networks (ICLR 2025 · similarity 0.89)
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.
