AkiraRust: Re-thinking LLM-aided Rust Repair Using a Feedback-guided Thinking Switch

2026-02-25

AI Summary

Existing approaches to leveraging large language models (LLMs) for repairing undefined behavior (UB) in Rust are hindered by rigid templates or the absence of executable semantics, making it difficult to simultaneously achieve context awareness and semantic correctness. This work proposes AkiraRust, a novel framework that integrates finite state machines with a multi-agent dual-mode reasoning mechanism. Guided by a waveform-driven controller, AkiraRust dynamically switches between fast and slow thinking modes at runtime to enable adaptive repair. The framework supports state rollback and semantic checkpoints, substantially enhancing both accuracy and robustness. Experimental results demonstrate that AkiraRust achieves a 92% semantic correctness rate while improving average repair speed by 2.2× over the current state-of-the-art method.

Technology Category

• Planning, Routing, and Scheduling: Replanning and Plan Repair
• Machine Learning: Large Multimodal Models (LMMs)
• Reasoning under Uncertainty: Uncertainty Representations

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Eliminating undefined behaviors (UBs) in Rust programs requires a deep semantic understanding to enable accurate and reliable repair. While existing studies have demonstrated the potential of LLMs to support Rust code analysis and repair, most frameworks remain constrained by inflexible templates or lack grounding in executable semantics, resulting in limited contextual awareness and semantic incorrectness. Here, we present AkiraRust, an LLM-driven repair and verification framework that incorporates a finite-state machine to dynamically adapt its detection and repair flow to runtime semantic conditions. AkiraRust introduces a dual-mode reasoning strategy that coordinates fast and slow thinking across multiple agents. Each agent is mapped to an FSM state, and a waveform-driven transition controller manages state switching, rollback decisions, and semantic check pointing, enabling context-aware and runtime-adaptive repair. Experimental results show that AkiraRust achieves about 92% semantic correctness and delivers a 2.2x average speedup compared to SOTA.

Problem

• undefined behaviors
• Rust repair
• semantic correctness
• LLM-aided
• context-awareness

Innovation

• LLM-driven repair
• finite-state machine
• dual-mode reasoning
• runtime-adaptive
• semantic correctness