Mol-Debate: Multi-Agent Debate Improves Structural Reasoning in Molecular Design
Wengyu Zhang, Xiao-Yong Wei, Qing Li · 2026
Read Mol-Debate: Multi-Agent Debate Improves Structural Reasoning in Molecular Design by Wengyu Zhang, Xiao-Yong Wei, Qing Li (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-20254.txt.
Introduction
Zhang, Wei, and Li present Mol-Debate, a debate driven loop for text guided molecular design, where word prompts meet strict chemical rules. Most rivals lean on one shot generation with fixed angles, while real drug work needs live multi angle critique and stepwise fixes. Mol-Debate runs a generate debate refine loop with perspective led orchestration, covering developer debater clashes, global local thought, and static dynamic fusion. It tops strong general and chemical baselines, with 59.82 percent exact match on ChEBI-20 and about 51 percent success on S2-Bench.
Catalog record
| CREATOR | Wengyu Zhang, Xiao-Yong Wei, Qing Li |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 50184 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.20254
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Mol-Debate: Multi-Agent Debate Improves Structural Reasoning in Molecular Design Wengyu Zhang, Xiao-Yong Wei, Qing Li Abstract Text-guided molecular design is a key capability for AI-driven drug discovery, yet it remains challenging to map sequential natural-language instructions with non-linear molecular structures under strict chemical constraints. Most existing approaches, including RAG, CoT prompting, and fine-tuning or RL, emphasize a small set of ad-hoc reasoning perspectives implemented in a largely one-shot generation pipeline. In contrast, real-world drug discovery relies on dynamic, multi-perspective critique and iterative refinement to reconcile semantic intent with structural feasibility. Motivated by this, we propose Mol-Debate , a generation paradigm that enables such dynamic reasoning through an iterative generate-debate-refine loop. We further characterize key challenges in this paradigm and address them through perspective-oriented orchestration, including developer-debater conflict, global-local structural reasoning, and static-dynamic integration. Experiments demonstrate that Mol-Debate achieves state-of-the-art performance against strong general and chemical baselines, reaching 59.82% exact match on ChEBI-20 and 50.52% weighted success rate on S 2 -Bench. Our code is available at https://github.com/wyuzh/Mol-Debate . † † Preprint. 1 Introduction Drug discovery is a focal point of AI for science, posing a critical multimodal NLP challenge despite the broad success of large language models (LLMs) 5 ; 1 ; 13 ; 42 . As shown in Figure 1 , the fundamental obstacle is a text-structure gap: translating sequential human instructions into the non-linear structural domain of molecules, which is defined by ring topology and functional group connectivity 14 . A range of methods have been proposed to enhance structural awareness in reasoning. These include attention-based alignment between language and molecular features 36 ; 32 ; Retrieval-Augmented Generation (RAG), which supplies relevant external chemical context 20 ; 22 ; and Chain-of-Thought (CoT) strategies that encourage structured reasoning 39 . Domain-specific models trained via fine-tuning (FT) and reinforcement learning (RL) on large chemistry corpora have also been developed to build foundational chemical competence 5 ; 31 ; 43 . Figure 1: Illustration of the text-structure gap: (a) sequential human instructions must be grounded in (b) non-linear structural domain of molecules. (1) The ring substructure is split across non-adjacent spans in the SMILES sequence. (2) The methoxy substructure appears as a single word in the text but corresponds to multiple atoms and bonds in the molecule. While effective, existing methods are often constrained to a few ad-hoc reasoning perspectives. For example, RAG mainly strengthens external knowledge retrieval, while pretraining and fine-tuning frameworks largely emphasize token alignment and data-driven pattern learning. As a result, these approaches are often tied to the specific perspective they are designed to exploit. Real-world drug discovery, however, demands a more integrative approach. The progress hinges on chemists’ ability to iteratively reconcile multiple viewpoints, such as high-level intent, structural constraints, and design trade-offs, through collaborative critique and refinement. While multi-agent systems (MAS) are a promising avenue for integrating the diverse perspectives required in drug design, their application raises a critical new issue 11 ; 35 . The system must navigate a delicate balance: preserving agent independence to avoid groupthink while ensuring enough coordination to reach a coherent decision 25 . If agents are too independent, their outputs fragment; if coordination is too strict, the system may converge prematurely and miss subtle errors. This new challenge positions Multi-Agent Debate (MAD) 3 as a compelling, yet unexplored, solution for molecular design, as its core mechanism of debate naturally balances independent reasoning with collaborative coordination. Motivated by this, we present a pilot study to investigate if debate-driven agentic reasoning can more effectively bridge the text-structure gap. To this end, we introduce Mol-Debate , an iterative generation framework (see Figure 2 ) where agents collaborate in a generate-debate-refine loop, integrating diverse and dynamic perspectives to improve structural fidelity. Mol-Debate addresses the integration of perspectives at three distinct levels. • Developer-Debater Conflict : A known issue where agents endowed with extensive domain expertise can experience a corresponding limitation in general reasoning and linguistic adaptability for debating 26 . To mitigate this asymmetry, we employ dedicated Debater Agents. Their function is to inject common-sense reasoning into the process, critically assessing how well candidate molecules satisfy the semantic intent of the prompt, thereby guaranteeing chemical robustness and precise user alignment. • Global-Local Structural Reasoning : Standard models typically process molecules as global, linear strings, which often causes them to overlook fine-grained chemical rules 5 ; 4 . Mol-Debate bridges this gap by enforcing a Local Structural Perspective. This perspective integrates a microscopic analysis of chemical constraints, shifting the model’s reasoning beyond mere “textual probability”. Consequently, the system can simultaneously respect both the global molecular syntax and the local chemical reality of its substructures. For example, this perspective helps Debater Agents prefer candidates with reasonable molecular weight and a reasonable number of rotatable bonds. • Static-Dynamic Integration : Current pipelines (e.g., RAG 22 , CoT 14 , and chemical LLMs 44 ) are fundamentally static, executing a one-way generation where errors are final. In contrast, Mol-Debate establishes a dynamic refinement system. It treats inter-agent disagreement not as failure, but as a learning signal. By creating a closed feedback loop, this consensus dissonance actively reformulates task constraints, allowing the system to refine its understanding of the query during inference, rather than passively accepting the initial output. Our contributions are threefold: • We identify the gap between prior methods that rely on a few ad-hoc reasoning perspectives and real-world drug discovery that requires dynamic, multi-perspective critique and refinement. • We propose Mol-Debate , a generation-by-debate paradigm for molecular design. • We characterize and address key challenges in this paradigm. Mol-Debate achieves the state-of-the-art performance and consistently high chemical validity against strong baselines. Figure 2: An overview of our Mol-Debate framework, an iterative generation framework where agents collaborate in a generate-debate-refine loop, integrating diverse and dynamic perspectives: (a) Developer-Debater Conflict, (b) Global-Local Structural Reasoning, and (c) Static-Dynamic Integration. 2 Related Work 2.1 Molecular Design Models Molecular design models connect natural language and molecular structures through retrieval, translation, editing, or optimization. Early work such as Text2Mol 6 learns joint text-molecule representations for caption-based retrieval, while later methods extend this line to bidirectional translation and in-context generation with LLMs 22 ; 23 ; 27 ; 28 . In parallel, chemical foundation models such as MolT5 5 , ChemDFM 44 , ChemDFM-R 43 , and Chem-R 37 improve text-level chemical competence through large-scale pretraining, reasoning supervision, distillation, and reinforcement learning. Other methods further strengthen the text-structure interface through explicit alignment or structural guidance, such as MolXPT 29 , GeLLM 3 O 2 , and MSR 14 . Recent work has also begun exploring agentic systems for molecular design. MT-Mol 18 uses multi-agent, tool-guided reasoning for molecular optimization, MultiMol 41 combines expert agents for multi-objective optimization, and PharmAgents 9 builds a broader multi-agent ecosystem for virtual pharmaceutical workflows. While these methods show the promise of agent-based molecular discovery, they mainly focus on optimization, editing, or broader drug-discovery pipelines. Our work is orthogonal to these model-centric advances: we treat such models as pluggable generators and focus on a debate-based framework that aims to orchestrate multiple models to better align text-structure intent with structural feasibility in molecular design. 2.2 Multi-Agent Debate Multi-Agent Debate (MAD) enhances robustness and reasoning diversity by allowing multiple agents to