Weighting What Matters: Boosting Sample Efficiency in Medical Report Generation via Token Reweighting
Researchers have developed a novel token reweighting technique to enhance the efficiency of training vision-language models (VLMs) for medical report generation. This method addresses the challenge of limited annotated data in the medical field by prioritizing semantically important tokens during training. Experiments demonstrated that this approach can achieve comparable report quality with up to ten times less training data, significantly improving sample efficiency. AI
IMPACT Improves data efficiency for medical report generation models, potentially reducing training costs and data requirements.