Research map / Optimization / Optimizers
Learning GFlowNets From Partial Episodes For Improved Convergence And Stability
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Train with SubTB(λ) instead of TB | GFlowNet · AMP generation | Top-100 Reward: 0.90 ± 0.03 -> 0.96 ± 0.02 | Helped |
| Train with SubTB(λ) instead of TB | GFlowNet · AMP generation | Top-100 Diversity: 31.42 ± 2.9 -> 42.23 ± 3.4 | Helped |
| Train with SubTB(λ) instead of TB | GFlowNet · GFP generation | Top-100 Reward: 0.76 ± 0.19 -> 1.18 ± 0.10 | Helped |
| Train with SubTB(λ) instead of TB | GFlowNet · GFP generation | Top-100 Diversity: 204.31 ± 0.44 -> 204.44 ± 0.45 | No clear effect |
| Use SubTB(λ) rather than TB | GFlowNet · Hypergrid | L1 distance | Helped |
| Use SubTB(λ) rather than TB | GFlowNet · Harder hypergrid with background reward 10^-4 | mode discovery | Helped |
| Use SubTB(λ) rather than TB, FM, and non-GFlowNet baselines | GFlowNet · Bit sequences | Spearman correlation | Helped |
| Use SubTB(λ) rather than TB, FM, DB, and non-GFlowNet baselines | GFlowNet · Antimicrobial peptide generation | Top-100 Reward | Helped |
| Use SubTB(λ) rather than TB, FM, DB, and non-GFlowNet baselines | GFlowNet · Antimicrobial peptide generation | Top-100 Diversity | Helped |
| Use intermediate λ values rather than the DB or TB limit (ablation) | SubTB(λ) · 8×8 hypergrid gradient experiment | gradient variance | Mixed |
| Train using only subtrajectories of length at most 4 (ablation) | SubTB(λ) · Hypergrid | L1 distance | No clear effect |
| Set λ to intermediate values (ablation) | SubTB(λ) · Inverse protein folding | Spearman correlation | Helped |
Closest papers by meaning
- Pre-Training and Fine-Tuning Generative Flow Networks (ICLR 2024 · similarity 0.93)
- GFlowNet Training by Policy Gradients (ICML 2024 · similarity 0.93)
- Flow Factorization for Efficient Generative Flow Networks (AAAI 2025 · similarity 0.93)
- On Divergence Measures for Training GFlowNets (NeurIPS 2024 · similarity 0.92)
- Optimizing Backward Policies in GFlowNets via Trajectory Likelihood Maximization (ICLR 2025 · similarity 0.92)
- Looking Backward: Retrospective Backward Synthesis for Goal-Conditioned GFlowNets (ICLR 2025 · similarity 0.92)
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.