MolLean: A Process-Verifiable Language for Chemical Reasoning
Abstract
MolLean introduces a verifiable chemical reasoning framework with explicit molecular states, Lean-based verification, and interactive tool feedback to reduce hallucinations across generation and reaction tasks.
Chemical chain-of-thought can contain hallucinated structural and property claims. We thereby introduce MolLean, a process-verifiable language that grounds chemical reasoning in an explicit molecular state. A layered Lean-based kernel checks state transitions and assertions over molecular readouts, from primitive graph edits to compound transformations. Meanwhile, a typed witness protocol separates native structural checks from externally supplied measurements. We further introduce MolLean-MCP, a stateful tool interface that exposes these transitions, checks, and readouts during generation: the model receives feedback to decide its next action, and rejected actions leave the state unchanged. MolLean supports one-pass trace execution, whole-trace refinement, and interactive reasoning through MolLean-MCP. We evaluate these three modes with direct chain-of-thought on 8,800 matched examples across 12 tasks spanning molecular generation, editing, optimization, and reactions. With Qwen3.6-27B, whole-trace refinement outperforms direct chain-of-thought on all 12 tasks. MolLean-MCP complements whole-trace refinement with stepwise verification feedback, reaching better performance in reaction tasks. Together, these results highlight an executable foundation for chemical reasoning, in which explicit chemical commitments can be checked and revised rather than remaining solely in free-form explanations.
Community
MolLean turns chemical reasoning into executable, checked steps. Direct chain-of-thought can miscount atoms, invent functional groups, or claim a property improved without measuring it. MolLean ties declared edits and claims to an explicit molecular state.
The video illustrates four examples from the paper: ethanol-to-1-propanol editing, an exact 25-carbon constraint, QED optimization with measured feedback, and an interactive MolLean-MCP episode where a rejected phenyl attachment leaves the molecule unchanged and guides the next successful action.
Across 8,800 matched examples covering 12 chemical task types, whole-trace refinement improves over direct CoT on all 12 with Qwen3.6-27B.
Paper: https://arxiv.org/abs/2610.20459
Project preview: https://github.com/phenixace/Open-MolLean
Code release is planned.
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- VETR: Execution-Grounded Self-Improving Agent for Symbolic Graphics Reasoning (2026)
- Constrained Program Generation for 3D Reaction Animation with a 0.8B Model (2026)
- TMCS: Tool-Grounded Multi-Agent Reasoning for Compositional Chemical Problem Solving (2026)
- Vestrum: Improving Agent Harnesses by Adapting Their Verification, Structure and Memory (2026)
- Agentic Interpretation: Harness-Governed Evidence Construction for Software Agents (2026)
- BTS-AgentBench: A Deterministic, Replayable Pipeline from Read-Only Telemetry Logs to Agent Benchmarks (2026)
- Beyond Answer Voting: Fine-Grained Correction of LLM Reasoning through Graph Alignment across Sampled Traces (2026)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: @librarian-bot recommend
Get this paper in your agent:
hf papers read 2610.20459 Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash Models citing this paper 0
No model linking this paper
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 0
No Space linking this paper
Collections including this paper 0
No Collection including this paper