Likun Zhao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Likun Zhao has a total of 1 co-authors. Here are the top co-authors by number of shared publications:
- Tao Zhang
Navigation college, Jimei University, Xiamen 361021, China
1 shared publication
Latest Publications
- Open Access
- 13 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 348-351
MMR: Math Multi-step Reasoning in Medical Dialogue Generation
With the advancements of large language models in text and visual tasks, researchers are increasingly exploring their applications in medical scenarios. While existing studies have successfully applied these models to medical dialogue systems, challenges remain in accurately calculating medication dosages and handling multi-turn reasoning due to the low tolerance for error in medical contexts. To address this, we propose Math Multi-step Reasoning in Medical Dialogue Generation (MMR), which enhances reasoning by iteratively breaking down complex problems into simpler questions using a “Least to Most Prompting” (LMP)strategy. MMR integrates Chain of Thought, React mechanisms, and Retrieval-Augmented Generation (RAG) with a domain-specific knowledge base to improve reasoning accuracy. Supported by the MedDGQA dataset, MMR outperforms state-of-the-art methods in both objective and subjective evaluations.
- 13 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 348-351
Open AccessMMR: Math Multi-step Reasoning in Medical Dialogue Generation
With the advancements of large language models in text and visual tasks, researchers are increasingly exploring their applications in medical scenarios. While existing studies have successfully applied these models to medical dialogue systems, challenges remain in accurately calculating medication dosages and handling multi-turn reasoning due to the low tolerance for error in medical contexts. To address this, we propose Math Multi-step Reasoning in Medical Dialogue Generation (MMR), which enhances reasoning by iteratively breaking down complex problems into simpler questions using a “Least to Most Prompting” (LMP)strategy. MMR integrates Chain of Thought, React mechanisms, and Retrieval-Augmented Generation (RAG) with a domain-specific knowledge base to improve reasoning accuracy. Supported by the MedDGQA dataset, MMR outperforms state-of-the-art methods in both objective and subjective evaluations.