# Towards Compositional Generalization in LLMs for Smart Contract Security: A Case Study on Reentrancy Vulnerabilities

arXiv: https://arxiv.org/abs/2601.06914 · HTML: https://mlsanity.com/paper/2601.06914/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 21 of its 40 references that are in the map.
- First posted on arXiv 11 Jan 2026.

## Builds on (most cited first, 21 of 21)

- [Finetuned Language Models Are Zero-Shot Learners](https://mlsanity.com/paper/2109.01652/) (2021)
- [The Implicit Bias of Gradient Descent on Separable Data](https://mlsanity.com/paper/1710.10345/) (2017)
- [Stronger generalization bounds for deep nets via a compression approach](https://mlsanity.com/paper/1802.05296/) (2018)
- [Unifying Molecular and Textual Representations via Multi-task Language Modelling](https://mlsanity.com/paper/2301.12586/) (2023)
- [Slither: A Static Analysis Framework For Smart Contracts](https://mlsanity.com/paper/1908.09878/) (2019)
- [Making Transformers Solve Compositional Tasks](https://mlsanity.com/paper/2108.04378/) (2021)
- [Securify: Practical Security Analysis of Smart Contracts](https://mlsanity.com/paper/1806.01143/) (2018)
- [Atom of Thoughts for Markov LLM Test-Time Scaling](https://mlsanity.com/paper/2502.12018/) (2025)
- [Compositional Generalization for Primitive Substitutions](https://mlsanity.com/paper/1910.02612/) (2019)
- [Jump to better conclusions: SCAN both left and right](https://mlsanity.com/paper/1809.04640/) (2018)
- [HeurAgenix: Leveraging LLMs for Solving Complex Combinatorial Optimization Challenges](https://mlsanity.com/paper/2506.15196/) (2025)
- [Complexity Control Facilitates Reasoning-Based Compositional Generalization in Transformers](https://mlsanity.com/paper/2501.08537/) (2025)
- [SmartBugs: A Framework to Analyze Solidity Smart Contracts](https://mlsanity.com/paper/2007.04771/) (2020)
- [Sereum: Protecting Existing Smart Contracts Against Re-Entrancy Attacks](https://mlsanity.com/paper/1812.05934/) (2018)
- [Reasoning-as-Logic-Units: Scaling Test-Time Reasoning in Large Language Models Through Logic Unit Alignment](https://mlsanity.com/paper/2502.07803/) (2025)
- [AtomR: Atomic Operator-Empowered Large Language Models for Heterogeneous Knowledge Reasoning](https://mlsanity.com/paper/2411.16495/) (2024)
- [From Chaos to Order: The Atomic Reasoner Framework for Fine-grained Reasoning in Large Language Models](https://mlsanity.com/paper/2503.15944/) (2025)
- [Robust Subtask Learning for Compositional Generalization](https://mlsanity.com/paper/2302.02984/) (2023)
- [Exploring Continual Learning of Compositional Generalization in NLI](https://mlsanity.com/paper/2403.04400/) (2024)
- [Compositional Security for Reentrant Applications](https://mlsanity.com/paper/2103.08577/) (2021)
- [Generative Large Language Model usage in Smart Contract Vulnerability Detection](https://mlsanity.com/paper/2504.04685/) (2025)

## Abstract (arXiv)

Large language models (LLMs) demonstrate remarkable capabilities in natural language understanding and generation. Despite being trained on large-scale, high-quality data, LLMs still fail to outperform traditional static analysis tools in specialized domains like smart contract vulnerability detection. To address this issue, this paper proposes a post-training algorithm based on atomic task decomposition and fusion. This algorithm aims to achieve combinatorial generalization under limited data by decomposing complex reasoning tasks. Specifically, we decompose the reentrancy vulnerability detection task into four linearly independent atomic tasks: identifying external calls, identifying state updates, identifying data dependencies between external calls and state updates, and determining their data flow order. These tasks form the core components of our approach. By training on synthetic datasets, we generate three compiler-verified datasets. We then employ the Slither tool to extract structural information from the control flow graph and data flow graph, which is used to fine-tune the LLM's adapter. Experimental results demonstrate that low-rank normalization fusion with the LoRA adapter improves the LLM's reentrancy vulnerability detection accuracy to 98.2%, surpassing state-of-the-art methods. On 31 real-world contracts, the algorithm achieves a 20% higher recall than traditional analysis tools.