Quantized Language Models in Healthcare: A Focus on Question Answering Based on the DPO Technique

• Mario Pinto Freitas Filho UFMA http://orcid.org/0000-0002-6294-685X
• João Dallyson Sousa de Almeida UFMA https://orcid.org/0000-0001-7013-9700
• Anselmo C. Paiva UFMA https://orcid.org/0000-0003-4921-0626

DOI: https://doi.org/10.5753/stil.2024.245458

Abstract

The agility in diagnosing patients is a vital factor for the skillful treatment of various ailments and is often the decisive parameter in patients' recovery. Considering that the average time spent by medical professionals on research activities is often around 4 hours, and recognizing the importance of increasingly reducing this estimate, this work seeks to explore the use of Large Language Models (LLMs) based on the Transformer architecture to optimize the time and efficiency of healthcare professionals' research activities. To achieve this, the objective is to understand what LLMs are through the Transformer and their functionalities, as well as to present the Medtext dataset used to train the model. Therefore, this work is an experimental study in which theoretical knowledge about LLMs and Transformers will be applied to address the issue and optimize research time.

Keywords: Applications of Natural Language Processing, Natural Language Tools and Resources, Question Answering, Spoken Language Processing

References

Chen, Z., Mao, H., Li, H., Jin, W., Wen, H., Wei, X., Wang, S., Yin, D., Fan, W., Liu, H., and Tang, J. (2023). Exploring the potential of large language models (llms) in learning on graphs.

Frantar, E., Ashkboos, S., Hoefler, T., and Alistarh, D. (2023). Gptq: Accurate post-training quantization for generative pre-trained transformers DOI: 10.48550/arXiv.2210.17323

Johnson, A. E. W., Stone, D. J., Celi, L. A., and Pollard, T. J. (2018). The mimic code repository: enabling reproducibility in critical care research. Journal of the American Medical Informatics Association, 25(1):32–39 DOI: 10.1093/jamia/ocx084

Li, X., Zhang, T., Dubois, Y., Taori, R., Gulrajani, I., Guestrin, C., Liang, P., and Hashimoto, T. B. (2023). Alpacaeval: An automatic evaluator of instruction-following models. [link] DOI: 10.48550/arXiv.2404.04475

Melamud, O. and Shivade, C. (2019). Towards automatic generation of shareable synthetic clinical notes using neural language models. In Proceedings of the 2nd Clinical Natural Language Processing Workshop at NAACL, pages 35–45 DOI: 10.48550/arXiv.1905.07002

Rafailov, R., Sharma, A., Mitchell, E., Ermon, S., Manning, C. D., and Finn, C. (2023). Direct preference optimization: Your language model is secretly a reward model DOI: 10.48550/arXiv.2305.18290

Russell-Rose T, C. J. (2017). Expert search strategies: The information retrieval practices of healthcare information professionals. JMIR Med Inform 2017;5(4):e33 DOI: 10.2196/medinform.7680

Touvron, H., Martin, L., and Al, E. (2023). Llama 2: Open foundation and fine-tuned chat models DOI: 10.48550/arXiv.2307.09288

Tunstall, L., Beeching, E., Lambert, N., Rajani, N., Rasul, K., Belkada, Y., Huang, S., von Werra, L., Fourrier, C., Habib, N., Sarrazin, N., Sanseviero, O., Rush, A. M., and Wolf, T. (2023). Zephyr: Direct distillation of lm alignment DOI: 10.48550/arXiv.2310.16944

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., and Polosukhin, I. (2023). Attention is all you need. DOI: 10.48550/arXiv.1706.03762

Zheng, L., Chiang, W.-L., Sheng, Y., Zhuang, S., Wu, Z., Zhuang, Y., Lin, Z., Li, Z., Li, D., Xing, E. P., Zhang, H., Gonzalez, J. E., and Stoica, I. (2023). Judging llm-as-a-judge with mt-bench and chatbot arena. DOI: 10.48550/arXiv.2306.05685