More about HKUST
Efficient Reasoning for Large Reasoning Model
PhD Qualifying Examination Title: "Efficient Reasoning for Large Reasoning Model" by Mr. Haoyue ZHANG Abstract: The rapid advancement of Large Language Models (LLMs) has significantly enhanced their capabilities in natural language understanding and complex reasoning tasks. However, the extensive use of Chain-of-Thought (CoT) prompting often leads to overly verbose reasoning processes, resulting in high computational costs and latency that hinder deployment in real-time, computation-sensitive applications. This paper provides a comprehensive survey on efficient reasoning methodologies for Large Reasoning Models (LRMs). In this paper, we introduce training-based approaches, including Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL), as well as innovative reasoning paradigms such as Prompt-guided efficient reasoning, switching of thinking, collaborative reasoning, and latent reasoning. Additionally, evaluation metrics and benchmarks tailored to assess reasoning efficiency are also examined. Finally, future research directions aimed at further optimizing reasoning processes in LRMs are discussed, including ensuring multi-modal application, scalability, reducing resource consumption, and addressing associated safety concerns. Date: Thursday, 14 August 2025 Time: 9:00am - 11:00am Venue: Room 3494 Lifts 25/26 Committee Members: Prof. Song Guo (Supervisor) Prof. Raymond Wong (Chairperson) Prof. Ke Yi