{"slug": "revisiting-llm-adaptation-for-3d-ct-report-generation-a-study-of-scaling-and", "title": "Revisiting LLM Adaptation for 3D CT Report Generation: A Study of Scaling and Diagnostic Priors", "summary": "Researchers introduced RAD3D-Prefix, a lightweight diagnostic-prior conditioning framework for volumetric CT report generation that keeps large language models frozen to minimize overfitting and clinical hallucination. A systematic study across LLMs from 96.1M to 1.6B parameters found that freezing larger models (~1B+) and training only lightweight projection layers offers the best trade-off between performance, generalization, and computational efficiency. RAD3D-Prefix outperformed comparable baselines in automatic metrics and a clinical reader study while using substantially fewer trainable parameters.", "body_md": "arXiv:2606.17213v1 Announce Type: new\nAbstract: Recent advances in multimodal learning, including large language models (LLMs) and vision-language models (VLMs), have demonstrated strong adaptability to natural images. However, extending their use to the medical domain, particularly for volumetric (3D) images, is challenging due to high computational complexity, volumetric dependencies and the semantic gap between visual features and clinical terminology. Naively fine-tuning LLMs on limited medical data often leads to overfitting and clinical hallucination, where linguistic fluency is prioritized over clinical factuality. In this study, we investigate parameter-efficient adaptation strategies for volumetric CT report generation and introduce RAD3D-Prefix, a lightweight diagnostic-prior conditioning framework that minimizes the need for extensive parameter training. This module integrates image embeddings with multi-label diagnostic classification logits, preserving critical clinical details while bridging the semantic gap. By keeping the LLM frozen, our method requires minimal trainable parameters and mitigates the risk of overfitting on small, domain-specific datasets. Through a systematic study spanning LLMs from 96.1M to 1.6B parameters, we find that fine-tuning is most beneficial for smaller LLMs, whereas freezing larger (~1B+ LLMs and training only lightweight projection layers provides a superior trade-off between performance, generalization, and computational efficiency. Across multiple automatic metrics and a clinical reader study, RAD3D-Prefix outperforms comparable parameter-efficient baselines and demonstrates strong out-of-domain generalization while using substantially fewer trainable parameters than fully fine-tuned alternatives.", "url": "https://wpnews.pro/news/revisiting-llm-adaptation-for-3d-ct-report-generation-a-study-of-scaling-and", "canonical_source": "https://arxiv.org/abs/2606.17213", "published_at": "2026-06-17 04:00:00+00:00", "updated_at": "2026-06-17 04:27:02.375603+00:00", "lang": "en", "topics": ["large-language-models", "computer-vision", "natural-language-processing", "ai-research", "ai-products"], "entities": ["RAD3D-Prefix", "LLM", "arXiv"], "alternates": {"html": "https://wpnews.pro/news/revisiting-llm-adaptation-for-3d-ct-report-generation-a-study-of-scaling-and", "markdown": "https://wpnews.pro/news/revisiting-llm-adaptation-for-3d-ct-report-generation-a-study-of-scaling-and.md", "text": "https://wpnews.pro/news/revisiting-llm-adaptation-for-3d-ct-report-generation-a-study-of-scaling-and.txt", "jsonld": "https://wpnews.pro/news/revisiting-llm-adaptation-for-3d-ct-report-generation-a-study-of-scaling-and.jsonld"}}