ARMOR: An Agentic Framework for Reaction Feasibility Prediction via Adaptive Utility-aware Multi-tool Reasoning
Authors: Ye Liu, Botao Yu, Xinyi Ling et al.
Summary
The authors built ARMOR, a system that decides whether a proposed chemical reaction will actually work by intelligently combining predictions from multiple AI tools. Instead of just averaging tool outputs or always trusting one tool, ARMOR learns which tools are reliable for which kinds of reactions, prioritizes the best tools, and uses memory-based reasoning to resolve cases where tools disagree. On a public chemistry dataset, this adaptive combination beats both single-tool approaches and simpler ways of aggregating multiple tools, with the biggest gains on reactions where tools give conflicting predictions.
Main takeaways:
- Reaction feasibility prediction tools (AI models for chemistry) vary widely in performance across different reactions, so no single tool is always best
- ARMOR organizes tools into a hierarchy that prioritizes top performers and defers to others when needed, rather than treating all tools equally
- Learns tool-specific patterns (when each tool is reliable) and uses memory-augmented reasoning to resolve conflicts
- Outperforms single-tool methods and simpler aggregation approaches, especially on reactions where tools disagree
- Code is available for replication
Relevance
Not related to my persona or midtraining work — this is a domain-specific chemistry application. Included because it's about combining multiple "tools" (models) intelligently, but the problem structure (conflicting predictions on chemistry) doesn't map to behavioral installation or persona detection.
Abstract
arXiv:2605.07103v1 Announce Type: new Abstract: Reaction feasibility prediction, as a fundamental problem in computational chemistry, has benefited from diverse tools enabled by recent advances in artificial intelligence, particularly large language models. However, the performance of individual tools varies substantially across reactions, making it difficult for any single tool to consistently perform well across all cases. This raises a critical challenge: how to effectively leverage multiple tools to obtain more accurate feasibility predictions. To address this, we propose ARMOR, an agentic framework that explicitly models tool-specific utilities, adaptively prioritizes tools, and further resolves the potential tool conflicts to produce the final prediction for each reaction. Unlike existing approaches that rely on simple aggregation or heuristic assignment over various tools, ARMOR organizes tools into a hierarchy that prioritizes top-performing tools and defers others when needed, characterizes their strengths through tool-specific patterns, and resolves conflicts via memoryaugmented reasoning. Extensive experiments on a public dataset demonstrate that ARMOR consistently outperforms strong baselines, including single-tool methods as well as various tool aggregation and tool selection approaches. Further analysis shows that the improvements are particularly significant on reactions with conflicting tool predictions, highlighting the effectiveness of ARMOR in leveraging the complementary strengths of multiple tools. The code is available via https://anonymous.4open.science/r/ARMOR-E13F.