pretty_name: JMID
language: ja
license: cc-by-4.0
task_categories:
- text-classification
- text-generation
size_categories:
- 1K<n<10K
configs:
- config_name: default
data_files:
- split: train
path: jmid.jsonl
- config_name: preview
data_files:
- split: train
path: jmid_preview.jsonl
JMID: Japanese Medical Incident Dataset
日本語
本データセットは、公益財団法人日本医療機能評価機構の医療事故報告書に書かれている医療事故内容から、医療事故の「具体的内容」「背景・要因」「改善策」とその他の情報をまとめたものである。
使い方の例は以下に載せる。
English
This dataset is compiled from the medical incident reports published by the Japan Council for Quality Health Care. It summarizes the contents of medical incidents, including the specific details, background and contributing factors, and proposed improvements, along with other related information.
An example of how to use the dataset is provided below.
from datasets import load_dataset
# Load the preview dataset (first 100 samples)
ds_preview = load_dataset("Sakaji-Lab/JMID", "preview")
print(ds_preview)
print(ds_preview["train"][0])
# Load the full dataset
# ds = load_dataset("Sakaji-Lab/JMID")
# print(ds) # or print()
# print(ds["train"][0])
日本語
医療事故の具体的内容、背景・要因、改善策のほかに、本データセットに含まれている追加情報は以下のようなものがある。
- 分類
- 事故の程度
- 専門分析班の議論
- 専門分析班・総合評価部会の議論
- 報告事例
- 記述情報
- 具体情報 etc...
English
In addition to the specific details, background and contributing factors, and proposed improvements of each medical incident, this dataset also includes the following supplementary information:
- Classification
- Severity of the incident
- Discussion by the expert analysis team
- Discussion by the expert analysis team and the General Evaluation Committee
- Reported cases
- Descriptive information
- Specific information etc...
医療事故内容から背景・要因、改善策を出力した時のBERTScoreの値
日本語 以下は、実際に本データセットを用い、医療事故の内容から、背景・要因、改善策を生成させたときの結果である。(第8回~第21回報告書の全2017件)
- Zero-shot
- All Prediction:事故内容を入力し、出力として背景・要因、改善策を一括生成させるタスク
- Each Prediction:事故内容を入力し、出力として背景・要因、改善策を段階的に生成させるタスク
- Few-shot
- All Prediction:事故内容を入力し、few-shot(n=5)で出力として背景・要因、改善策を一括生成させるタスク
- Each Prediction:事故内容を入力し、few-shot(n=5)で出力として背景・要因、改善策を段階的に生成させるタスク
English
The following shows the results obtained using this dataset to generate background and contributing factors as well as proposed improvements from the descriptions of medical incidents (a total of 2,017 cases from the 8th to the 21st reports):
Zero-shot
- All Prediction: A task in which the incident description is given as input,and both the background/contributing factors and proposed improvements are generated at once.
- Each Prediction: A task in which the incident description is given as input, and the background/contributing factors and proposed improvements are generated step by step.
Few-shot
- All Prediction: A task in which the incident description is given as input, and both the background/contributing factors and proposed improvements are generated at once using few-shot learning (n=5).
- Each Prediction: A task in which the incident description is given as input, and the background/contributing factors and proposed improvements are generated step by step using few-shot learning (n=5).
医療事故の内容から、背景・要因を生成させた時のBERTScoreの平均値 (Average BERTScore when generating background and contributing factors from the description of medical incidents)
| モデル | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.620 | 0.698 | 0.656 | 0.577 | 0.689 | 0.628 | 0.707 | 0.717 | 0.711 | 0.695 | 0.717 | 0.705 |
| Mistral | 0.600 | 0.691 | 0.642 | 0.566 | 0.684 | 0.619 | 0.622 | 0.661 | 0.640 | 0.656 | 0.699 | 0.676 |
| Llama3.3 | 0.612 | 0.698 | 0.651 | 0.591 | 0.697 | 0.639 | 0.686 | 0.712 | 0.698 | 0.671 | 0.705 | 0.687 |
| MedSwallow | 0.678 | 0.636 | 0.655 | 0.666 | 0.626 | 0.644 | 0.715 | 0.689 | 0.700 | 0.673 | 0.648 | 0.659 |
| JMedLoRA | 0.679 | 0.657 | 0.667 | 0.680 | 0.720 | 0.698 | 0.712 | 0.686 | 0.698 | 0.714 | 0.683 | 0.697 |
医療事故の内容から、改善策を生成させた時のBERTScoreの平均値 (Average BERTScore when generating proposed improvements from the description of medical incidents)
| モデル | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.609 | 0.698 | 0.650 | 0.593 | 0.696 | 0.640 | 0.692 | 0.730 | 0.710 | 0.679 | 0.727 | 0.702 |
| Mistral | 0.601 | 0.694 | 0.644 | 0.650 | 0.684 | 0.666 | 0.656 | 0.681 | 0.668 | 0.649 | 0.719 | 0.682 |
| Llama3.3 | 0.615 | 0.700 | 0.654 | 0.592 | 0.696 | 0.640 | 0.691 | 0.731 | 0.710 | 0.688 | 0.724 | 0.705 |
| MedSwallow | 0.027 | 0.025 | 0.026 | 0.667 | 0.616 | 0.639 | 0.647 | 0.617 | 0.631 | 0.685 | 0.651 | 0.667 |
| JMedLoRA | 0.539 | 0.519 | 0.528 | 0.468 | 0.450 | 0.459 | 0.703 | 0.688 | 0.695 | 0.723 | 0.705 | 0.713 |
具体情報を用いた際の、医療事故内容別の結果は以下である。
The following shows the results for each type of medical incident when utilizing the specific information.
背景・要因(background and contributing factors)
調剤(71件)
| 調剤 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.630 | 0.690 | 0.658 | 0.677 | 0.677 | 0.697 | 0.711 | 0.693 | 0.702 | 0.693 | 0.692 | 0.692 |
| Mistral | 0.614 | 0.683 | 0.646 | 0.681 | 0.681 | 0.629 | 0.649 | 0.661 | 0.655 | 0.553 | 0.565 | 0.558 |
| Llama3.3 | 0.624 | 0.688 | 0.654 | 0.685 | 0.685 | 0.639 | 0.697 | 0.690 | 0.693 | 0.675 | 0.688 | 0.681 |
| MedSwallow | 0.681 | 0.625 | 0.651 | 0.566 | 0.566 | 0.594 | 0.714 | 0.674 | 0.693 | 0.649 | 0.610 | 0.628 |
| JMedLoRA | 0.689 | 0.642 | 0.664 | 0.709 | 0.709 | 0.699 | 0.709 | 0.658 | 0.682 | 0.720 | 0.660 | 0.688 |
薬剤(438件)
| 薬剤 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.624 | 0.699 | 0.659 | 0.692 | 0.692 | 0.699 | 0.713 | 0.715 | 0.713 | 0.697 | 0.712 | 0.704 |
| Mistral | 0.608 | 0.692 | 0.646 | 0.689 | 0.689 | 0.627 | 0.624 | 0.653 | 0.638 | 0.571 | 0.597 | 0.583 |
| Llama3.3 | 0.617 | 0.696 | 0.654 | 0.696 | 0.696 | 0.642 | 0.689 | 0.705 | 0.696 | 0.673 | 0.699 | 0.685 |
| MedSwallow | 0.673 | 0.629 | 0.650 | 0.612 | 0.612 | 0.633 | 0.715 | 0.684 | 0.698 | 0.676 | 0.649 | 0.661 |
| JMedLoRA | 0.662 | 0.632 | 0.646 | 0.722 | 0.722 | 0.698 | 0.710 | 0.679 | 0.693 | 0.710 | 0.669 | 0.688 |
人工呼吸器(182件)
| 人工呼吸器(182件) | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.601 | 0.69 | 0.641 | 0.687 | 0.687 | 0.69 | 0.698 | 0.718 | 0.707 | 0.68 | 0.712 | 0.695 |
| Mistral | 0.583 | 0.684 | 0.629 | 0.675 | 0.675 | 0.607 | 0.611 | 0.664 | 0.636 | 0.528 | 0.566 | 0.546 |
| Llama3.3 | 0.601 | 0.695 | 0.644 | 0.693 | 0.693 | 0.632 | 0.675 | 0.712 | 0.692 | 0.66 | 0.702 | 0.679 |
| MedSwallow | 0.68 | 0.636 | 0.656 | 0.664 | 0.664 | 0.681 | 0.719 | 0.693 | 0.704 | 0.694 | 0.669 | 0.68 |
| JMedLoRA | 0.683 | 0.672 | 0.676 | 0.720 | 0.720 | 0.697 | 0.708 | 0.69 | 0.698 | 0.712 | 0.686 | 0.698 |
輸液ポンプ(114件)
| 輸液ポンプ(114件) | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.629 | 0.703 | 0.663 | 0.695 | 0.695 | 0.707 | 0.721 | 0.718 | 0.719 | 0.703 | 0.715 | 0.708 |
| Mistral | 0.611 | 0.692 | 0.648 | 0.692 | 0.692 | 0.632 | 0.604 | 0.639 | 0.620 | 0.617 | 0.641 | 0.628 |
| Llama3.3 | 0.621 | 0.700 | 0.657 | 0.696 | 0.696 | 0.643 | 0.698 | 0.716 | 0.706 | 0.681 | 0.704 | 0.691 |
| MedSwallow | 0.686 | 0.642 | 0.663 | 0.664 | 0.664 | 0.691 | 0.710 | 0.675 | 0.691 | 0.675 | 0.646 | 0.659 |
| JMedLoRA | 0.671 | 0.638 | 0.653 | 0.719 | 0.719 | 0.700 | 0.720 | 0.680 | 0.699 | 0.716 | 0.674 | 0.693 |
持参薬(136件)
| 持参薬 (136件) |
Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.646 | 0.713 | 0.678 | 0.709 | 0.709 | 0.716 | 0.725 | 0.721 | 0.723 | 0.72 | 0.726 | 0.722 |
| Mistral | 0.631 | 0.709 | 0.667 | 0.703 | 0.703 | 0.645 | 0.656 | 0.68 | 0.667 | 0.638 | 0.665 | 0.651 |
| Llama3.3 | 0.635 | 0.71 | 0.67 | 0.707 | 0.707 | 0.656 | 0.707 | 0.718 | 0.712 | 0.695 | 0.714 | 0.704 |
| MedSwallow | 0.68 | 0.641 | 0.659 | 0.61 | 0.61 | 0.632 | 0.725 | 0.693 | 0.708 | 0.669 | 0.638 | 0.652 |
| JMedLoRA | 0.699 | 0.666 | 0.681 | 0.73 | 0.73 | 0.711 | 0.721 | 0.684 | 0.702 | 0.723 | 0.683 | 0.702 |
ドレーンの挿入・管理(134件)
| ドレーンの挿入・管理 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.626 | 0.699 | 0.66 | 0.693 | 0.693 | 0.704 | 0.713 | 0.717 | 0.714 | 0.704 | 0.719 | 0.711 |
| Mistral | 0.605 | 0.691 | 0.645 | 0.681 | 0.681 | 0.619 | 0.617 | 0.653 | 0.634 | 0.567 | 0.592 | 0.579 |
| Llama3.3 | 0.616 | 0.701 | 0.656 | 0.702 | 0.702 | 0.643 | 0.693 | 0.714 | 0.703 | 0.671 | 0.703 | 0.686 |
| MedSwallow | 0.687 | 0.643 | 0.664 | 0.609 | 0.609 | 0.633 | 0.731 | 0.686 | 0.707 | 0.705 | 0.662 | 0.682 |
| JMedLoRA | 0.685 | 0.661 | 0.672 | 0.721 | 0.721 | 0.703 | 0.719 | 0.684 | 0.7 | 0.714 | 0.677 | 0.694 |
(臨床)検査(84件)
| (臨床)検査 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.59 | 0.677 | 0.63 | 0.679 | 0.679 | 0.678 | 0.703 | 0.725 | 0.713 | 0.685 | 0.72 | 0.701 |
| Mistral | 0.57 | 0.668 | 0.615 | 0.654 | 0.654 | 0.587 | 0.635 | 0.687 | 0.659 | 0.532 | 0.566 | 0.548 |
| Llama3.3 | 0.601 | 0.698 | 0.645 | 0.697 | 0.697 | 0.63 | 0.681 | 0.716 | 0.697 | 0.665 | 0.708 | 0.685 |
| MedSwallow | 0.674 | 0.637 | 0.654 | 0.603 | 0.603 | 0.614 | 0.71 | 0.69 | 0.698 | 0.677 | 0.652 | 0.663 |
| JMedLoRA | 0.682 | 0.665 | 0.672 | 0.722 | 0.722 | 0.696 | 0.715 | 0.69 | 0.701 | 0.712 | 0.686 | 0.698 |
小児患者の療養(133件)
| 小児患者の療養 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.617 | 0.7 | 0.656 | 0.69 | 0.69 | 0.694 | 0.714 | 0.735 | 0.723 | 0.701 | 0.731 | 0.715 |
| Mistral | 0.595 | 0.694 | 0.641 | 0.686 | 0.686 | 0.614 | 0.633 | 0.684 | 0.657 | 0.611 | 0.656 | 0.632 |
| Llama3.3 | 0.611 | 0.705 | 0.654 | 0.704 | 0.704 | 0.639 | 0.684 | 0.722 | 0.702 | 0.674 | 0.716 | 0.694 |
| MedSwallow | 0.678 | 0.637 | 0.656 | 0.646 | 0.646 | 0.664 | 0.726 | 0.703 | 0.714 | 0.682 | 0.663 | 0.671 |
| JMedLoRA | 0.702 | 0.686 | 0.693 | 0.726 | 0.726 | 0.704 | 0.722 | 0.698 | 0.709 | 0.724 | 0.698 | 0.71 |
リハビリテーション(131件)
| リハビリテーション | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.636 | 0.712 | 0.671 | 0.704 | 0.704 | 0.71 | 0.718 | 0.733 | 0.724 | 0.708 | 0.733 | 0.72 |
| Mistral | 0.611 | 0.707 | 0.655 | 0.703 | 0.703 | 0.634 | 0.632 | 0.677 | 0.653 | 0.647 | 0.69 | 0.667 |
| Llama3.3 | 0.619 | 0.709 | 0.66 | 0.707 | 0.707 | 0.645 | 0.692 | 0.727 | 0.709 | 0.68 | 0.722 | 0.7 |
| MedSwallow | 0.664 | 0.623 | 0.642 | 0.656 | 0.656 | 0.674 | 0.715 | 0.695 | 0.704 | 0.681 | 0.658 | 0.668 |
| JMedLoRA | 0.685 | 0.668 | 0.675 | 0.717 | 0.717 | 0.696 | 0.711 | 0.693 | 0.701 | 0.703 | 0.687 | 0.694 |
輸血療法(121件)
| 輸血療法 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.628 | 0.708 | 0.665 | 0.694 | 0.694 | 0.695 | 0.712 | 0.724 | 0.717 | 0.701 | 0.722 | 0.711 |
| Mistral | 0.605 | 0.697 | 0.647 | 0.695 | 0.695 | 0.628 | 0.634 | 0.671 | 0.652 | 0.596 | 0.635 | 0.614 |
| Llama3.3 | 0.608 | 0.691 | 0.646 | 0.701 | 0.701 | 0.642 | 0.691 | 0.718 | 0.704 | 0.675 | 0.711 | 0.692 |
| MedSwallow | 0.694 | 0.657 | 0.674 | 0.634 | 0.634 | 0.648 | 0.707 | 0.688 | 0.697 | 0.648 | 0.635 | 0.64 |
| JMedLoRA | 0.692 | 0.673 | 0.681 | 0.725 | 0.725 | 0.705 | 0.717 | 0.695 | 0.705 | 0.72 | 0.689 | 0.703 |
禁忌薬(83件)
| 禁忌薬 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.633 | 0.712 | 0.67 | 0.705 | 0.705 | 0.711 | 0.718 | 0.728 | 0.722 | 0.704 | 0.73 | 0.716 |
| Mistral | 0.615 | 0.707 | 0.658 | 0.701 | 0.701 | 0.634 | 0.62 | 0.659 | 0.638 | 0.611 | 0.651 | 0.63 |
| Llama3.3 | 0.623 | 0.709 | 0.662 | 0.705 | 0.705 | 0.64 | 0.688 | 0.717 | 0.701 | 0.678 | 0.715 | 0.695 |
| MedSwallow | 0.664 | 0.621 | 0.641 | 0.621 | 0.621 | 0.64 | 0.72 | 0.697 | 0.707 | 0.658 | 0.631 | 0.643 |
| JMedLoRA | 0.667 | 0.644 | 0.655 | 0.721 | 0.721 | 0.699 | 0.713 | 0.693 | 0.702 | 0.709 | 0.676 | 0.691 |
放射線検査(71件)
| 放射線検査 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.594 | 0.697 | 0.64 | 0.7 | 0.7 | 0.688 | 0.675 | 0.713 | 0.692 | 0.663 | 0.716 | 0.687 |
| Mistral | 0.576 | 0.693 | 0.628 | 0.679 | 0.679 | 0.599 | 0.59 | 0.652 | 0.618 | 0.527 | 0.581 | 0.552 |
| Llama3.3 | 0.572 | 0.681 | 0.621 | 0.683 | 0.683 | 0.613 | 0.654 | 0.705 | 0.678 | 0.64 | 0.7 | 0.668 |
| MedSwallow | 0.665 | 0.633 | 0.647 | 0.631 | 0.631 | 0.635 | 0.688 | 0.696 | 0.69 | 0.658 | 0.663 | 0.659 |
| JMedLoRA | 0.665 | 0.669 | 0.666 | 0.71 | 0.71 | 0.672 | 0.685 | 0.691 | 0.687 | 0.676 | 0.672 | 0.672 |
左右(部位)間違い(73件)
| 左右(部位)間違い | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.584 | 0.674 | 0.625 | 0.672 | 0.672 | 0.671 | 0.681 | 0.708 | 0.693 | 0.664 | 0.703 | 0.682 |
| Mistral | 0.561 | 0.665 | 0.607 | 0.656 | 0.656 | 0.584 | 0.6 | 0.65 | 0.623 | 0.536 | 0.585 | 0.558 |
| Llama3.3 | 0.581 | 0.677 | 0.624 | 0.671 | 0.671 | 0.606 | 0.65 | 0.695 | 0.67 | 0.639 | 0.69 | 0.662 |
| MedSwallow | 0.662 | 0.646 | 0.653 | 0.641 | 0.641 | 0.647 | 0.687 | 0.679 | 0.681 | 0.617 | 0.615 | 0.614 |
| JMedLoRA | 0.681 | 0.67 | 0.674 | 0.71 | 0.71 | 0.687 | 0.695 | 0.683 | 0.687 | 0.691 | 0.682 | 0.685 |
異物残存(23件)
| 異物残存 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.631 | 0.703 | 0.664 | 0.694 | 0.694 | 0.705 | 0.716 | 0.721 | 0.718 | 0.691 | 0.709 | 0.7 |
| Mistral | 0.618 | 0.701 | 0.656 | 0.662 | 0.662 | 0.605 | 0.609 | 0.642 | 0.625 | 0.532 | 0.56 | 0.545 |
| Llama3.3 | 0.621 | 0.698 | 0.657 | 0.7 | 0.7 | 0.657 | 0.685 | 0.707 | 0.695 | 0.68 | 0.71 | 0.694 |
| MedSwallow | 0.701 | 0.666 | 0.683 | 0.639 | 0.639 | 0.656 | 0.707 | 0.676 | 0.691 | 0.573 | 0.563 | 0.567 |
| JMedLoRA | 0.679 | 0.658 | 0.667 | 0.718 | 0.718 | 0.699 | 0.71 | 0.681 | 0.695 | 0.706 | 0.682 | 0.694 |
ベッドなど病室の設備(71件)
| ベッドなど病室の設備 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.626 | 0.707 | 0.663 | 0.68 | 0.68 | 0.682 | 0.705 | 0.727 | 0.715 | 0.691 | 0.726 | 0.707 |
| Mistral | 0.597 | 0.699 | 0.643 | 0.69 | 0.69 | 0.617 | 0.633 | 0.683 | 0.656 | 0.651 | 0.701 | 0.674 |
| Llama3.3 | 0.599 | 0.696 | 0.643 | 0.693 | 0.693 | 0.632 | 0.675 | 0.715 | 0.694 | 0.669 | 0.713 | 0.69 |
| MedSwallow | 0.685 | 0.639 | 0.661 | 0.619 | 0.619 | 0.637 | 0.724 | 0.699 | 0.71 | 0.681 | 0.648 | 0.663 |
| JMedLoRA | 0.681 | 0.677 | 0.678 | 0.711 | 0.711 | 0.693 | 0.71 | 0.692 | 0.7 | 0.708 | 0.691 | 0.699 |
化学療法(60件)
| 化学療法 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.625 | 0.696 | 0.658 | 0.696 | 0.696 | 0.703 | 0.708 | 0.714 | 0.711 | 0.693 | 0.709 | 0.7 |
| Mistral | 0.606 | 0.69 | 0.645 | 0.687 | 0.687 | 0.625 | 0.632 | 0.668 | 0.649 | 0.518 | 0.546 | 0.531 |
| Llama3.3 | 0.616 | 0.694 | 0.652 | 0.691 | 0.691 | 0.638 | 0.685 | 0.705 | 0.695 | 0.675 | 0.704 | 0.688 |
| MedSwallow | 0.668 | 0.621 | 0.643 | 0.571 | 0.571 | 0.593 | 0.707 | 0.676 | 0.69 | 0.678 | 0.653 | 0.665 |
| JMedLoRA | 0.667 | 0.637 | 0.651 | 0.714 | 0.714 | 0.691 | 0.71 | 0.676 | 0.692 | 0.713 | 0.673 | 0.692 |
電気メス等(39件)
| 電気メス等 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.563 | 0.645 | 0.6 | 0.671 | 0.671 | 0.681 | 0.688 | 0.712 | 0.698 | 0.672 | 0.705 | 0.698 |
| Mistral | 0.525 | 0.619 | 0.567 | 0.567 | 0.567 | 0.508 | 0.603 | 0.654 | 0.627 | 0.571 | 0.613 | 0.591 |
| Llama3.3 | 0.607 | 0.7 | 0.649 | 0.695 | 0.695 | 0.64 | 0.663 | 0.711 | 0.685 | 0.651 | 0.703 | 0.675 |
| MedSwallow | 0.678 | 0.635 | 0.654 | 0.606 | 0.606 | 0.623 | 0.712 | 0.691 | 0.7 | 0.632 | 0.616 | 0.623 |
| JMedLoRA | 0.671 | 0.657 | 0.663 | 0.713 | 0.713 | 0.695 | 0.705 | 0.69 | 0.696 | 0.711 | 0.688 | 0.698 |
患者取り違え(53件)
| 患者取り違え | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.629 | 0.704 | 0.664 | 0.698 | 0.698 | 0.704 | 0.71 | 0.715 | 0.712 | 0.696 | 0.718 | 0.706 |
| Mistral | 0.602 | 0.694 | 0.644 | 0.692 | 0.692 | 0.626 | 0.61 | 0.646 | 0.627 | 0.521 | 0.549 | 0.534 |
| Llama3.3 | 0.605 | 0.697 | 0.647 | 0.693 | 0.693 | 0.632 | 0.693 | 0.713 | 0.702 | 0.674 | 0.706 | 0.689 |
| MedSwallow | 0.698 | 0.672 | 0.684 | 0.631 | 0.631 | 0.643 | 0.714 | 0.692 | 0.702 | 0.706 | 0.685 | 0.695 |
| JMedLoRA | 0.692 | 0.67 | 0.68 | 0.724 | 0.724 | 0.703 | 0.714 | 0.689 | 0.701 | 0.717 | 0.684 | 0.699 |
改善策(proposed improvements)
調剤(71件)
| 調剤 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.626 | 0.699 | 0.660 | 0.713 | 0.713 | 0.699 | 0.705 | 0.724 | 0.714 | 0.689 | 0.721 | 0.704 |
| Mistral | 0.619 | 0.693 | 0.654 | 0.681 | 0.681 | 0.669 | 0.675 | 0.689 | 0.681 | 0.325 | 0.341 | 0.333 |
| Llama3.3 | 0.631 | 0.700 | 0.664 | 0.692 | 0.692 | 0.645 | 0.707 | 0.722 | 0.714 | 0.703 | 0.718 | 0.710 |
| MedSwallow | 0.020 | 0.019 | 0.019 | 0.564 | 0.564 | 0.596 | 0.675 | 0.634 | 0.653 | 0.719 | 0.666 | 0.691 |
| JMedLoRA | 0.553 | 0.521 | 0.536 | 0.448 | 0.448 | 0.464 | 0.681 | 0.646 | 0.662 | 0.736 | 0.695 | 0.714 |
薬剤(438件)
| 薬剤 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.618 | 0.702 | 0.657 | 0.717 | 0.717 | 0.697 | 0.701 | 0.731 | 0.715 | 0.685 | 0.726 | 0.705 |
| Mistral | 0.612 | 0.697 | 0.652 | 0.687 | 0.687 | 0.667 | 0.659 | 0.675 | 0.666 | 0.513 | 0.557 | 0.534 |
| Llama3.3 | 0.623 | 0.699 | 0.659 | 0.695 | 0.695 | 0.643 | 0.7 | 0.728 | 0.713 | 0.694 | 0.72 | 0.706 |
| MedSwallow | 0.022 | 0.019 | 0.02 | 0.6 | 0.6 | 0.628 | 0.643 | 0.606 | 0.623 | 0.724 | 0.681 | 0.701 |
| JMedLoRA | 0.504 | 0.478 | 0.49 | 0.406 | 0.406 | 0.416 | 0.703 | 0.679 | 0.69 | 0.723 | 0.693 | 0.707 |
人工呼吸器(182件)
| 人工呼吸器(182件) | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.603 | 0.692 | 0.644 | 0.706 | 0.706 | 0.681 | 0.688 | 0.729 | 0.707 | 0.677 | 0.723 | 0.698 |
| Mistral | 0.598 | 0.692 | 0.641 | 0.682 | 0.682 | 0.669 | 0.655 | 0.682 | 0.667 | 0.494 | 0.545 | 0.517 |
| Llama3.3 | 0.613 | 0.698 | 0.652 | 0.695 | 0.695 | 0.639 | 0.686 | 0.728 | 0.706 | 0.679 | 0.715 | 0.696 |
| MedSwallow | 0.042 | 0.040 | 0.041 | 0.644 | 0.644 | 0.670 | 0.659 | 0.630 | 0.643 | 0.727 | 0.692 | 0.708 |
| JMedLoRA | 0.537 | 0.521 | 0.528 | 0.399 | 0.399 | 0.407 | 0.711 | 0.695 | 0.702 | 0.713 | 0.692 | 0.701 |
輸液ポンプ(114件)
| 輸液ポンプ(114件) | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.617 | 0.706 | 0.659 | 0.723 | 0.723 | 0.700 | 0.697 | 0.733 | 0.714 | 0.683 | 0.731 | 0.706 |
| Mistral | 0.609 | 0.703 | 0.652 | 0.695 | 0.695 | 0.674 | 0.625 | 0.653 | 0.638 | 0.391 | 0.422 | 0.405 |
| Llama3.3 | 0.620 | 0.707 | 0.660 | 0.700 | 0.700 | 0.644 | 0.694 | 0.734 | 0.713 | 0.694 | 0.728 | 0.710 |
| MedSwallow | 0.032 | 0.029 | 0.031 | 0.663 | 0.663 | 0.689 | 0.585 | 0.554 | 0.568 | 0.720 | 0.686 | 0.702 |
| JMedLoRA | 0.501 | 0.480 | 0.490 | 0.417 | 0.417 | 0.428 | 0.709 | 0.687 | 0.697 | 0.721 | 0.701 | 0.710 |
持参薬(136件)
| 持参薬 (136件) |
Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.618 | 0.711 | 0.661 | 0.729 | 0.729 | 0.702 | 0.701 | 0.743 | 0.721 | 0.686 | 0.737 | 0.710 |
| Mistral | 0.612 | 0.706 | 0.656 | 0.696 | 0.696 | 0.675 | 0.669 | 0.698 | 0.683 | 0.437 | 0.471 | 0.453 |
| Llama3.3 | 0.621 | 0.707 | 0.661 | 0.703 | 0.703 | 0.647 | 0.703 | 0.744 | 0.722 | 0.704 | 0.740 | 0.721 |
| MedSwallow | 0.044 | 0.040 | 0.042 | 0.615 | 0.615 | 0.635 | 0.679 | 0.648 | 0.662 | 0.728 | 0.697 | 0.712 |
| JMedLoRA | 0.527 | 0.507 | 0.517 | 0.512 | 0.512 | 0.520 | 0.707 | 0.691 | 0.698 | 0.739 | 0.721 | 0.729 |
ドレーンの挿入・管理(134件)
| ドレーンの挿入・管理 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.612 | 0.697 | 0.651 | 0.714 | 0.714 | 0.690 | 0.694 | 0.730 | 0.711 | 0.682 | 0.727 | 0.704 |
| Mistral | 0.602 | 0.694 | 0.644 | 0.684 | 0.684 | 0.668 | 0.649 | 0.674 | 0.660 | 0.414 | 0.444 | 0.428 |
| Llama3.3 | 0.616 | 0.701 | 0.655 | 0.699 | 0.699 | 0.643 | 0.692 | 0.732 | 0.711 | 0.684 | 0.717 | 0.700 |
| MedSwallow | 0.049 | 0.046 | 0.047 | 0.598 | 0.598 | 0.623 | 0.677 | 0.643 | 0.659 | 0.726 | 0.686 | 0.705 |
| JMedLoRA | 0.562 | 0.544 | 0.553 | 0.433 | 0.433 | 0.441 | 0.719 | 0.701 | 0.709 | 0.712 | 0.727 | 0.701 |
(臨床)検査(84件)
| (臨床)検査 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.582 | 0.674 | 0.624 | 0.699 | 0.699 | 0.670 | 0.687 | 0.735 | 0.709 | 0.670 | 0.725 | 0.696 |
| Mistral | 0.578 | 0.676 | 0.622 | 0.670 | 0.670 | 0.651 | 0.673 | 0.707 | 0.689 | 0.375 | 0.405 | 0.389 |
| Llama3.3 | 0.609 | 0.707 | 0.654 | 0.698 | 0.698 | 0.637 | 0.686 | 0.734 | 0.709 | 0.683 | 0.723 | 0.702 |
| MedSwallow | 0.009 | 0.009 | 0.009 | 0.602 | 0.602 | 0.621 | 0.671 | 0.641 | 0.654 | 0.729 | 0.693 | 0.710 |
| JMedLoRA | 0.589 | 0.568 | 0.577 | 0.543 | 0.543 | 0.549 | 0.713 | 0.704 | 0.708 | 0.720 | 0.711 | 0.715 |
小児患者の療養(133件)
| 小児患者の療養 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.604 | 0.693 | 0.645 | 0.700 | 0.700 | 0.673 | 0.688 | 0.730 | 0.708 | 0.674 | 0.726 | 0.699 |
| Mistral | 0.592 | 0.689 | 0.636 | 0.670 | 0.670 | 0.656 | 0.674 | 0.696 | 0.684 | 0.511 | 0.555 | 0.532 |
| Llama3.3 | 0.614 | 0.702 | 0.654 | 0.695 | 0.695 | 0.636 | 0.688 | 0.732 | 0.708 | 0.687 | 0.724 | 0.704 |
| MedSwallow | 0.026 | 0.024 | 0.025 | 0.624 | 0.624 | 0.647 | 0.659 | 0.624 | 0.640 | 0.727 | 0.697 | 0.711 |
| JMedLoRA | 0.585 | 0.564 | 0.574 | 0.500 | 0.500 | 0.508 | 0.695 | 0.690 | 0.692 | 0.705 | 0.696 | 0.700 |
リハビリテーション(131件)
| リハビリテーション | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.623 | 0.713 | 0.664 | 0.721 | 0.721 | 0.694 | 0.696 | 0.739 | 0.717 | 0.684 | 0.736 | 0.709 |
| Mistral | 0.608 | 0.707 | 0.654 | 0.690 | 0.690 | 0.673 | 0.661 | 0.686 | 0.673 | 0.604 | 0.666 | 0.633 |
| Llama3.3 | 0.621 | 0.711 | 0.663 | 0.706 | 0.706 | 0.645 | 0.686 | 0.739 | 0.711 | 0.689 | 0.732 | 0.709 |
| MedSwallow | 0.036 | 0.035 | 0.035 | 0.642 | 0.642 | 0.664 | 0.625 | 0.601 | 0.612 | 0.731 | 0.699 | 0.714 |
| JMedLoRA | 0.546 | 0.529 | 0.537 | 0.438 | 0.438 | 0.444 | 0.687 | 0.685 | 0.685 | 0.707 | 0.702 | 0.704 |
輸血療法(121件)
| 輸血療法 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.611 | 0.706 | 0.654 | 0.704 | 0.704 | 0.678 | 0.692 | 0.738 | 0.713 | 0.678 | 0.731 | 0.703 |
| Mistral | 0.601 | 0.697 | 0.645 | 0.691 | 0.691 | 0.673 | 0.660 | 0.685 | 0.671 | 0.572 | 0.640 | 0.603 |
| Llama3.3 | 0.601 | 0.689 | 0.641 | 0.697 | 0.697 | 0.637 | 0.691 | 0.737 | 0.713 | 0.688 | 0.733 | 0.709 |
| MedSwallow | 0.022 | 0.022 | 0.022 | 0.611 | 0.611 | 0.631 | 0.613 | 0.592 | 0.602 | 0.721 | 0.687 | 0.702 |
| JMedLoRA | 0.574 | 0.552 | 0.562 | 0.497 | 0.497 | 0.505 | 0.695 | 0.682 | 0.688 | 0.731 | 0.715 | 0.722 |
禁忌薬(83件)
| 禁忌薬 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.613 | 0.700 | 0.653 | 0.718 | 0.718 | 0.692 | 0.691 | 0.729 | 0.709 | 0.678 | 0.727 | 0.701 |
| Mistral | 0.606 | 0.695 | 0.647 | 0.691 | 0.691 | 0.673 | 0.658 | 0.681 | 0.669 | 0.550 | 0.612 | 0.579 |
| Llama3.3 | 0.619 | 0.701 | 0.657 | 0.696 | 0.696 | 0.640 | 0.689 | 0.728 | 0.707 | 0.683 | 0.719 | 0.700 |
| MedSwallow | 0.028 | 0.024 | 0.026 | 0.602 | 0.602 | 0.622 | 0.665 | 0.635 | 0.649 | 0.714 | 0.682 | 0.696 |
| JMedLoRA | 0.487 | 0.473 | 0.479 | 0.464 | 0.464 | 0.475 | 0.696 | 0.680 | 0.687 | 0.714 | 0.694 | 0.703 |
放射線検査(71件)
| 放射線検査 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.590 | 0.688 | 0.635 | 0.711 | 0.711 | 0.677 | 0.670 | 0.725 | 0.696 | 0.655 | 0.716 | 0.683 |
| Mistral | 0.581 | 0.687 | 0.629 | 0.692 | 0.692 | 0.664 | 0.637 | 0.676 | 0.655 | 0.626 | 0.711 | 0.665 |
| Llama3.3 | 0.580 | 0.677 | 0.623 | 0.686 | 0.686 | 0.623 | 0.663 | 0.722 | 0.690 | 0.661 | 0.711 | 0.684 |
| MedSwallow | 0.032 | 0.030 | 0.031 | 0.622 | 0.622 | 0.638 | 0.599 | 0.581 | 0.589 | 0.706 | 0.693 | 0.698 |
| JMedLoRA | 0.557 | 0.543 | 0.549 | 0.528 | 0.528 | 0.533 | 0.708 | 0.707 | 0.706 | 0.694 | 0.696 | 0.694 |
左右(部位)間違い(73件)
| 左右(部位)間違い | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.579 | 0.685 | 0.627 | 0.703 | 0.703 | 0.672 | 0.681 | 0.735 | 0.706 | 0.665 | 0.725 | 0.693 |
| Mistral | 0.574 | 0.681 | 0.622 | 0.675 | 0.675 | 0.657 | 0.650 | 0.690 | 0.669 | 0.630 | 0.718 | 0.670 |
| Llama3.3 | 0.592 | 0.694 | 0.639 | 0.687 | 0.687 | 0.623 | 0.669 | 0.734 | 0.699 | 0.674 | 0.727 | 0.699 |
| MedSwallow | 0.010 | 0.009 | 0.009 | 0.641 | 0.641 | 0.658 | 0.600 | 0.586 | 0.592 | 0.723 | 0.704 | 0.713 |
| JMedLoRA | 0.528 | 0.516 | 0.521 | 0.452 | 0.452 | 0.455 | 0.683 | 0.682 | 0.681 | 0.724 | 0.717 | 0.719 |
異物残存(23件)
| 異物残存 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.618 | 0.703 | 0.657 | 0.719 | 0.719 | 0.695 | 0.700 | 0.727 | 0.713 | 0.681 | 0.721 | 0.700 |
| Mistral | 0.615 | 0.706 | 0.657 | 0.667 | 0.667 | 0.652 | 0.653 | 0.675 | 0.663 | 0.653 | 0.718 | 0.683 |
| Llama3.3 | 0.618 | 0.697 | 0.655 | 0.702 | 0.702 | 0.647 | 0.691 | 0.730 | 0.709 | 0.692 | 0.726 | 0.708 |
| MedSwallow | 0 | 0 | 0 | 0.631 | 0.631 | 0.659 | 0.673 | 0.626 | 0.648 | 0.707 | 0.691 | 0.698 |
| JMedLoRA | 0.553 | 0.528 | 0.540 | 0.354 | 0.354 | 0.359 | 0.692 | 0.667 | 0.679 | 0.731 | 0.701 | 0.715 |
ベッドなど病室の設備(71件)
| ベッドなど病室の設備 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.608 | 0.699 | 0.650 | 0.689 | 0.689 | 0.658 | 0.681 | 0.732 | 0.705 | 0.668 | 0.727 | 0.696 |
| Mistral | 0.591 | 0.693 | 0.637 | 0.688 | 0.688 | 0.674 | 0.672 | 0.703 | 0.686 | 0.642 | 0.717 | 0.676 |
| Llama3.3 | 0.607 | 0.696 | 0.647 | 0.694 | 0.694 | 0.632 | 0.676 | 0.727 | 0.700 | 0.681 | 0.724 | 0.701 |
| MedSwallow | 0.008 | 0.008 | 0.008 | 0.620 | 0.620 | 0.641 | 0.695 | 0.675 | 0.683 | 0.740 | 0.701 | 0.719 |
| JMedLoRA | 0.579 | 0.569 | 0.573 | 0.506 | 0.506 | 0.512 | 0.705 | 0.708 | 0.705 | 0.711 | 0.710 | 0.710 |
化学療法(60件)
| 化学療法 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.613 | 0.701 | 0.654 | 0.721 | 0.721 | 0.695 | 0.690 | 0.727 | 0.708 | 0.676 | 0.721 | 0.698 |
| Mistral | 0.606 | 0.695 | 0.647 | 0.683 | 0.683 | 0.662 | 0.658 | 0.682 | 0.669 | 0.641 | 0.713 | 0.675 |
| Llama3.3 | 0.616 | 0.697 | 0.654 | 0.694 | 0.694 | 0.639 | 0.691 | 0.733 | 0.711 | 0.689 | 0.721 | 0.704 |
| MedSwallow | 0.014 | 0.011 | 0.012 | 0.568 | 0.568 | 0.592 | 0.665 | 0.632 | 0.648 | 0.705 | 0.678 | 0.690 |
| JMedLoRA | 0.487 | 0.469 | 0.477 | 0.396 | 0.396 | 0.405 | 0.703 | 0.687 | 0.694 | 0.727 | 0.699 | 0.712 |
電気メス等(39件)
| 電気メス等 | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.552 | 0.635 | 0.590 | 0.687 | 0.687 | 0.664 | 0.670 | 0.711 | 0.689 | 0.659 | 0.708 | 0.682 |
| Mistral | 0.528 | 0.613 | 0.567 | 0.558 | 0.558 | 0.551 | 0.654 | 0.677 | 0.665 | 0.627 | 0.705 | 0.663 |
| Llama3.3 | 0.603 | 0.689 | 0.643 | 0.686 | 0.686 | 0.628 | 0.664 | 0.716 | 0.688 | 0.668 | 0.714 | 0.689 |
| MedSwallow | 0.034 | 0.032 | 0.033 | 0.602 | 0.602 | 0.625 | 0.702 | 0.668 | 0.684 | 0.709 | 0.696 | 0.701 |
| JMedLoRA | 0.558 | 0.540 | 0.549 | 0.410 | 0.410 | 0.416 | 0.701 | 0.686 | 0.692 | 0.709 | 0.692 | 0.699 |
患者取り違え(53件)
| 患者取り違え | Zero-shot (All Prediction) |
Zero-shot (Each Prediction) |
Few-shot (All Prediction) |
Few-shot (Each Prediction) |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
| GPT-4o | 0.619 | 0.721 | 0.666 | 0.741 | 0.741 | 0.712 | 0.703 | 0.751 | 0.725 | 0.692 | 0.747 | 0.718 |
| Mistral | 0.605 | 0.711 | 0.653 | 0.715 | 0.715 | 0.693 | 0.646 | 0.676 | 0.660 | 0.666 | 0.738 | 0.700 |
| Llama3.3 | 0.611 | 0.710 | 0.656 | 0.705 | 0.705 | 0.642 | 0.705 | 0.749 | 0.726 | 0.702 | 0.743 | 0.722 |
| MedSwallow | 0.014 | 0.013 | 0.014 | 0.644 | 0.644 | 0.663 | 0.679 | 0.646 | 0.661 | 0.757 | 0.711 | 0.732 |
| JMedLoRA | 0.632 | 0.617 | 0.624 | 0.542 | 0.542 | 0.550 | 0.743 | 0.725 | 0.733 | 0.739 | 0.726 | 0.732 |