MolHop: Scaffold-Hopping Molecular Design with Agentic Reinforcement Learning
We propose MolHop, an agentic framework for cross-scaffold molecular discovery that enables models to navigate and edit within a pre-organized chemical space.
ProofPaper ↗
Key points
- Molecular design is often formulated as an optimization problem over existing molecular spaces.
- On drug molecule and commercial compound tasks, connecting Grok-4.6 and DeepSeek-Flash-V4.1 to the MolHop framework substantially improves their cross-scaffold joint success rates.
- We develop MolHop-9B through agentic reinforcement learning, and MolHop-9B outperforms generative baselines including GPT-5.6-sol, Claude-5-Opus, and Gemini-3.1-Pro, while generating candidate molecules with larger scaffold differences from the reference molecules.
- These results demonstrate that effective molecular discovery beyond the traditional optimization paradigm requires structured chemical and action spaces, enabling agents to explore new structural regions rather than merely refine existing molecules.
Sources (1)
- [1]MolHop: Scaffold-Hopping Molecular Design with Agentic Reinforcement LearningarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 14, 10:11 AM
We propose MolHop, an agentic framework for cross-scaffold molecular discovery that enables models to navigate and edit within a pre-organized chemical space.
Molecular design is often formulated as an optimization problem over existing molecular spaces.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 13, 2026PMO-Dock: Benchmarking Docking, Specificity, and Generalization in Molecular Optimization
- Oct 13, 2026Build long-horizon agents with Grok 4.7 on Amazon Bedrock
- Oct 13, 2026TSRL: Trigger-based Supervised Reinforcement Learning for Continual Agent Training
- Oct 13, 2026Urbex: Agentic Spatial Grounding for City-Scale 3D Scenes
- Oct 13, 2026Tev: Scaling Long-Horizon Agentic Environments with Dense Progress Rewards
- Oct 12, 2026llama.cpp optimizations for specific hardware and models