Research map / Reinforcement learning / Visual and representation RL
GRASP: Navigating Retrosynthetic Planning with Goal-driven Policy
Yemin Yu, Ying Wei, Kun Kuang, Zhengxing Huang, Huaxiu Yao, Fei Wu
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 GRASP planner instead of Retro* | MT single-step planning · Pistachio | success rate: 0.92 -> 0.95 | Improved |
| Use GRASP planner instead of Retro* | MT single-step planning · WuxiTest | success rate: 0.80 -> 0.86 | Improved |
| Use GRASP planner instead of Retro* | MT single-step planning · Pistachio | avg. time: 62.0 -> 42.6 | Improved |
| Use GRASP planner instead of Retro* | MT single-step planning · WuxiTest | avg. time: 112.5 -> 79.3 | Improved |
| Use GRASP instead of HgSearch on average cost | MT single-step planning · Pistachio | avg. cost: 7.06 -> 7.47 | Hurt |
| Use GRASP instead of HgSearch on average cost | MT single-step planning · WuxiTest | avg. cost: 21.41 -> 21.19 | Improved |
| Use GRASP instead of Retro* for pathway length | MT single-step planning · WuxiTest | avg. length: 7.50 -> 6.93 | Improved |
| Use GRASP instead of Retro* for pathway length | MT single-step planning · Pistachio | avg. length: 4.27 -> 4.12 | Improved |
| Specify expert goals for goal-driven planning instead of general planning | GRASP · WuxiTest user study | avg. rating (0-10): 8.3 -> 9.2 | Improved |
| Use GRASP goal-driven planning with MCTS source building blocks | GRASP · WuxiTest user study | avg. rating (0-10): 6.5 -> 7.5 | Improved |
| Replace Retro* with GRASP in self-improved retrosynthetic planning | Self-improved planning · Pistachio, WuxiTest | success rate | Improved |
| Add self-imitation learning (SIL) (ablation) | GRASP training · Pistachio | training success rate | Improved |
Later papers that compared with this method
1 later accepted paper reports GRASP as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- BRIDGE: Bi-level Reinforcement Learning for Dynamic Group Structure in Coalition Formation Games (ICLR 2026 · better on 1 of 1 shared result)
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
- LLM-Augmented Chemical Synthesis and Design Decision Programs (ICML 2025 · similarity 0.90)
- Retrosynthesis Planning via Worst-path Policy Optimisation in Tree-structured MDPs (NeurIPS 2025 · similarity 0.90)
- Double-Ended Synthesis Planning with Goal-Constrained Bidirectional Search (NeurIPS 2024 · similarity 0.90)
- Retro-fallback: retrosynthetic planning in an uncertain world (ICLR 2024 · similarity 0.89)
- RetroBridge: Modeling Retrosynthesis with Markov Bridges (ICLR 2024 · similarity 0.89)
- URSA: Chemistry-Aware Benchmark for Utilitarian Retrosynthesis Assessment (NeurIPS 2026 · 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.
