Research map / Probabilistic and scientific modeling / Molecules and proteins
Variational Learning for Insertion-based Generation
Yangtian Zhang, Zhe Wang, Arthur Gretton, Rex Ying, David van Dijk, Michalis Titsias, Jiaxin Shi
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 |
|---|---|---|---|
| learn the insertion policy instead of random insertion order (AO-IP) (ablation) | IP · GuacaMol unconditional | validity (%): 88.2 -> 97.2 | Improved |
| learn the insertion policy instead of random insertion order (AO-IP) (ablation) | IP · GuacaMol unconditional | valid-unique-novel (%): 87.7 -> 94.9 | Improved |
| learn the insertion policy (ablation) | IP · Fragment decoration | validity (%): 40.9 -> 99.8 | Improved |
| learn the insertion policy (ablation) | IP · Linker design | validity (%): 50.6 -> 99.5 | Improved |
| learn the insertion policy (ablation) | IP · Linker + full decoration | unique valid (%): 31.1 -> 34.2 | Improved |
| use classifier-based instead of policy-based termination | IP · GuacaMol unconditional | valid-unique-novel (%): 94.9 -> 95.6 | Improved |
| insertion-based IP instead of fixed left-to-right FO-ARM | IP · Star graph planning | sequence accuracy (%): 24.0 -> 83.0 | Improved |
| learned insertion order IP instead of LO-ARM | IP · Star graph planning | sequence accuracy (%): 30.1 -> 83.0 | Improved |
| IP instead of FO-ARM on distribution matching | IP · GuacaMol unconditional | validity (%): 98.5 -> 97.2 | Hurt |
| IP instead of FO-ARM on novelty | IP · GuacaMol unconditional | valid-unique-novel (%): 80.4 -> 94.9 | Improved |
| relax insertion slot constraints (add decoration slots) | IP · Conditional molecule generation | unique valid (%): 13.1 -> 34.2 | Improved |
Closest papers by meaning
- Learning-Order Autoregressive Models with Application to Molecular Graph Generation (ICML 2025 · similarity 0.85)
- Unlocking Guidance for Discrete State-Space Diffusion and Flow Models (ICLR 2025 · similarity 0.84)
- Train for the Worst, Plan for the Best: Understanding Token Ordering in Masked Diffusions (ICML 2025 · similarity 0.84)
- Controlled Generation with Equivariant Variational Flow Matching (ICML 2025 · similarity 0.84)
- Derivative-Free Guidance in Continuous and Discrete Diffusion Models with Soft Value-based Decoding (NeurIPS 2025 · similarity 0.83)
- Absorb & Escape: Overcoming Single Model Limitations in Generating Heterogeneous Genomic Sequences (NeurIPS 2024 · similarity 0.83)
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.
