Research map / Learning theory and algorithms / Combinatorial and search optimization
From Feasible to Practical: Pareto-Optimal Synthesis Planning
Friedrich Hastedt, Dongda Zhang, Antonio Del rio chanona
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 multi-objective MORetro* with BO weight sampling instead of single-objective Retro* | Retro* with G2E · USPTO-190 | HV: 0.98 -> 1.04 | Improved |
| use MORetro* (BO) instead of Retro* | Retro* with PDVN · USPTO-190 | HV: 0.88 -> 0.92 | Improved |
| use MORetro* (BO) instead of Retro* | Retro* with NeuralSym · USPTO-190 | HV: 0.52 -> 0.56 | Improved |
| use adaptive BO weight sampling instead of fixed weight vector [.2,.2,.2,.4] (ablation) | Fixed scalarization with PDVN · USPTO-190 | success rate (%): 69.8 -> 96.8 | Improved |
| use adaptive BO weight sampling instead of fixed weights (ablation) | Fixed scalarization with G2E · USPTO-190 | success rate (%): 90.5 -> 97.9 | Improved |
| use adaptive BO weight sampling instead of fixed weights (ablation) | Fixed scalarization with NeuralSym · USPTO-190 | success rate (%): 47.4 -> 60.0 | Improved |
| use MORetro* (BO) instead of single-objective Retro* | Retro* · USPTO-190 | success rate (%): 100 -> 97.9 (G2E); 98.9 -> 96.8 (PDVN); 61.6 -> 60.0 (NeuralSym) | Hurt |
| use MORetro* (BO) instead of Retro* | Retro* · USPTO-190 | Max. DS: 0.80 -> 0.84 (G2E); 0.72 -> 0.76 (PDVN); 0.57 -> 0.64 (NeuralSym) | Improved |
| use BO sampling instead of Sobol sampling (ablation) | MORetro* · USPTO-190 | PF quality, success rate, diversity | Improved |
| use BO sampling instead of grid sampling (ablation) | MORetro* · USPTO-190 | HV, success rate | Mixed |
| add exact V_bound pruning of open nodes (ablation) | MORetro* (BO) · USPTO-190 G2E | search space and PF quality: average search space reduction 54%; HV +0.005; R2 -0.03 | Improved |
| relax pruning to eps-dominance with eps=0.1 (ablation) | V_bound pruning · USPTO-190 G2E | search space reduction: 54% -> 61% average reduction; model calls 289 -> 265 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Retro* — Retro*: Learning Retrosynthetic Planning with Neural Guided A* Search (ICML 2020) · better on 11 of 13 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Double-Ended Synthesis Planning with Goal-Constrained Bidirectional Search (NeurIPS 2024 · similarity 0.91)
- Retrosynthesis Planning via Worst-path Policy Optimisation in Tree-structured MDPs (NeurIPS 2025 · similarity 0.89)
- Retro-fallback: retrosynthetic planning in an uncertain world (ICLR 2024 · similarity 0.89)
- LLM-Augmented Chemical Synthesis and Design Decision Programs (ICML 2025 · similarity 0.88)
- URSA: Chemistry-Aware Benchmark for Utilitarian Retrosynthesis Assessment (NeurIPS 2026 · similarity 0.85)
- RetroBridge: Modeling Retrosynthesis with Markov Bridges (ICLR 2024 · similarity 0.85)
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.
